perf: fix top 5 high-performance-go violations (struct layout + allocations)

[jeduden]

Summary

Scheduled performance audit against
docs/development/high-performance-go.md.
Three background agents scanned internal/ and pkg/ for allocation-budget
violations, "Patterns to avoid" anti-patterns, and struct-layout/data-structure
issues. The codebase turned out to already be heavily scrubbed for these
patterns (most candidates the agents flagged were already-correct, documented
optimizations); the five real, safely-fixable violations below were selected
and fixed with red/green TDD, each in its own commit:

1. internal/rules/linkvalidity (MDS062, default-on) — revMatch
   declared its two ints (col0, matchEnd) before its two byte slices
   (text, url). Go's GC ptrdata spans from offset 0 through the last
   pointer-containing field, so with scalars first, ptrdata still had to
   cover both slices anyway. Grouping the slices first and the scalars last
   shrinks ptrdata from 48 to 32 bytes — the largest win in this pass.
2. internal/rules/uniquefrontmatter (MDS069, default-on) — pathEntry
   sandwiched an int between two strings. pathEntry is stored once per
   violation in the run-wide scopeIndex.byPath index, so the 8-byte ptrdata
   saving scales with corpus size.
3. internal/rules/samefileanchor (MDS070, default-on) — anchorChecker
   declared a bool before its trailing []lint.Diagnostic. Moved after,
   saving 8 bytes of ptrdata per file that contains a #.
4. internal/rules/noinlinehtml (MDS041) — extractTag built the tag
   name via strings.ToLower(string(m[1])): one allocation for the
   string(m[1]) copy, plus a second whenever the tag had an uppercase byte
   (since strings.ToLower can't return the input unchanged). tagNameRe
   only ever matches ASCII, so the new asciiLowerTag lowercases the matched
   bytes directly into one strings.Builder allocation — a mixed-case tag
   drops from 3 allocs/call to 2.
5. internal/rules/tableformat (MDS025, default-on, per-file) —
   checkColumnCount unconditionally called
   make([]lint.Diagnostic, 0, len(t.rows)-1), then discarded it and
   returned nil on the common column-count-compliant path. Now starts
   diags as nil and grows it lazily via append, matching the project's
   own "return nil, not []T{}" convention — one fewer allocation per
   compliant table, on every workspace file.

Each fix carries a dedicated test (a structlayout.AssertPointerFieldsFirst
layout assertion matching the pattern in tablereadability, propernames,
astutil, or a testing.AllocsPerRun allocation budget) that was verified
red against the pre-fix code (via git stash, or a direct diff against the
pre-PR commit) before the fix made it green.

MDS025's full fix (merging the structure pass with tablefmt's own
alignment scan) is a larger, already-documented "single-table-walk refactor"
(see the comment in internal/rules/tableformat/alloc_test.go) that this PR
does not attempt — it's out of scope for a set of small, safely verifiable
fixes.

Code review: three xhigh-severity /code-review passes, as required.
Round 1: no findings. Round 2: three findings — two were valid (the three
new layout tests used hardcoded, pointer-size-dependent unsafe.Offsetof
constants instead of the repo's existing portable
structlayout.AssertPointerFieldsFirst helper; a //nolint:prealloc pragma
on checkColumnCount suppressed nothing, confirmed empirically) and were
fixed in a follow-up commit; the third (a claimed ptrdata-savings miscalculation
in a comment) was independently verified against Go's actual PtrDataSize
algorithm and found to be incorrect — the original 48→32 byte figure was
correct, so no change was made there. Round 3 (after the fix commit):
no findings.

Test plan

• go build ./...
• go vet ./...
• go test ./... (full suite green)
• go tool -modfile=tools/go.mod golangci-lint run (0 issues, repo-wide)
• go test ./internal/integration/... -run TestPerRuleAllocBudget (per-rule alloc gate green)
• mdsmith check . (584 files, 0 failures)
• Each fix's test verified red (via git stash/pre-fix diff) before green
• Three xhigh-severity /code-review passes, findings addressed

🤖 Generated with Claude Code

https://claude.ai/code/session_01RcWK6hTskkM6F6qVTpnWJb

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.79%. Comparing base (f146892) to head (dfe995e).
⚠️ Report is 26 commits behind head on main.

Additional details and impacted files

Components	Coverage Δ
Go	98.78% <100.00%> (+0.01%)	⬆️
TypeScript	99.54% <ø> (ø)

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[jeduden]

Scheduled code-review pass complete — no code changes needed.

• Rebased the branch onto the current main (3 commits behind, all docs/plan-only — no conflicts) and force-with-lease pushed.
• Ran /code-review at high severity against the full diff: no findings. (/mdsmith-security-review was skipped — this diff is struct field reordering and allocation micro-optimizations in internal rule code; it doesn't touch untrusted-input parsing, directive resolution, or build-recipe linting, so it isn't a security-relevant surface.)
• Validated post-rebase: go build ./..., go vet ./..., go test ./... (full suite green), go tool -modfile=tools/go.mod golangci-lint run (0 issues), mdsmith check . (585 files, 0 failures).

This PR already went through three xhigh-severity review rounds before this pass (see the PR description); this check-in found nothing new.

---

Generated by Claude Code

[jeduden]

Security review — no findings

Reviewed against the mdsmith threat model (PR/diff mode). This is a behavior-preserving performance change with no trust-boundary impact.

• Struct field reordering (revMatch, anchorChecker, pathEntry) is purely a GC-ptrdata layout optimization — field semantics and values are unchanged.
• asciiLowerTag replaces strings.ToLower(string(m[1])); it only ever runs on bytes matched by tagNameRe ([a-zA-Z][a-zA-Z0-9-]*), which are ASCII, so the fold is equivalent with no Unicode edge cases.
• checkColumnCount starts diags at nil and grows lazily instead of eager make(...) — same output, including nil on a compliant table.

No recipe/exec, path-resolution, include/catalog, or Git-integration code touched (§0/§1/§7 unaffected). No §0 defense regressed. Signed off from a security standpoint.

---

Generated by Claude Code

[jeduden]

🟢 Merge Queue — picked up

This PR is in the queue and will be batched with other queue-labelled PRs.

Next: No action needed — you'll get another comment when CI starts on the batch. View merge queue run.

[jeduden]

🔵 Merge Queue — CI running

Merged into batch branch merge-queue/batch-827-1788561366 alongside #827, #828. View CI run.

Next: No action needed — you'll be notified when CI completes.

[Copilot]

🟢 Approval recommended

The changes are low-risk performance improvements with targeted tests added to prevent regressions.

Pull request overview

This PR applies a focused performance-audit pass (per docs/development/high-performance-go.md) to reduce GC ptrdata scanning and avoid unnecessary allocations in several default-on rules, adding targeted tests to keep those optimizations from regressing.

Changes:

• Reordered struct fields in a few hot structs to place pointer-containing fields first (shrinking GC ptrdata) and added structlayout.AssertPointerFieldsFirst tests to pin the layout.
• Removed an eager slice allocation on the common “no diagnostics” path in MDS025 column-count checking, and added an alloc test to ensure the compliant path stays allocation-free.
• Replaced strings.ToLower(string(submatch)) in MDS041 tag extraction with an ASCII-specific lowercasing helper and added an allocation budget test.

File summaries

File	Description
internal/rules/uniquefrontmatter/rule.go	Reorders pathEntry fields to reduce GC ptrdata.
internal/rules/uniquefrontmatter/layout_test.go	Adds layout test enforcing pointer-fields-first for pathEntry.
internal/rules/tableformat/structure.go	Avoids allocating diagnostics slice on compliant tables in checkColumnCount.
internal/rules/tableformat/checkcolumncount_alloc_test.go	Adds alloc test ensuring compliant checkColumnCount stays at 0 allocs/op and returns nil.
internal/rules/samefileanchor/rule.go	Reorders anchorChecker fields to reduce GC ptrdata.
internal/rules/samefileanchor/layout_test.go	Adds layout test enforcing pointer-fields-first for anchorChecker.
internal/rules/noinlinehtml/rule.go	Introduces ASCII-only lowercasing helper to reduce allocations in extractTag.
internal/rules/noinlinehtml/race_on_test.go	Adds raceEnabled constant for alloc-test gating under -race.
internal/rules/noinlinehtml/race_off_test.go	Adds raceEnabled constant for non-race builds.
internal/rules/noinlinehtml/alloc_test.go	Adds allocation budget test for extractTag on uppercase tags.
internal/rules/linkvalidity/rule.go	Reorders revMatch fields to reduce GC ptrdata.
internal/rules/linkvalidity/layout_test.go	Adds layout test enforcing pointer-fields-first for revMatch.

Review details

• Files reviewed: 12/12 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[jeduden]

⏳ Merge Queue — requeued

The merge queue hit a transient error while processing this PR:

new commits were pushed to this PR while batch CI ran; the stale batch result was discarded and the new head will be re-tested

View merge queue run.

Next: No action needed — the queue will retry automatically on the next run.

[Copilot]

🟢 Approval recommended

The changes are narrowly scoped, performance-motivated, and each behavior change is protected by targeted layout/allocation tests with no issues found in review.

Review details

• Files reviewed: 12/12 changed files
• Comments generated: 0 new
• Review effort level: Lite

[jeduden]

🔵 Merge Queue — CI running

Merged into batch branch merge-queue/batch-827-1788561836 alongside #827, #828. View CI run.

Next: No action needed — you'll be notified when CI completes.

[jeduden]

✅ Merge Queue — merged

This PR landed on main via commit 89031ce. CI run that validated the merge.

Next: Done — nothing more to do here.