perf: add comprehensive benchmark suite with CI regression detection

[dlevy-msft-sql]

Problem

The driver has no systematic benchmark coverage and no CI guardrails against performance regressions. Changes that degrade hot-path performance can merge undetected.

Solution

Add 96 benchmarks covering all core driver hot paths, with automated CI regression detection using benchstat.

Benchmarks Added

Category	Count	File
Token parsing	10	token_benchmark_test.go
TDS buffer I/O	10	buf_benchmark_test.go
String encoding	10	tds_benchmark_test.go
Type conversion	13	convert_benchmark_test.go
Wire protocol types	17	types_benchmark_test.go
RPC encoding	7	rpc_benchmark_test.go
Bulk copy params	9	bulkcopy_benchmark_test.go
Integration round-trips	14	integration_benchmark_test.go
Connection string parsing	5	msdsn/conn_str_benchmark_test.go

CI Integration

• Dedicated benchmarks job in GitHub Actions with SQL Server 2025 in Docker
• Warmup run (-benchtime=100ms -count=1) stabilizes CPU caches and turbo boost before measurement; results are discarded
• Runs baseline benchmarks on main via git worktree, then PR benchmarks, both with -count=10 -benchtime=1s
• Compares results using benchstat -alpha=0.01 (p < 0.01 significance threshold)
• Fails CI on statistically significant regressions exceeding 15% magnitude
• Reports improvements via ::notice annotations
• Explicit benchmark pattern avoids pre-existing slow/problematic benchmarks
• Copies benchmark test files to main worktree for fair baseline comparison
• Excludes TdsBuffer_Write_Large from regression gate (cache-sensitive ~120ns operation with 30-46% natural variance)

Bug Fix

Fixes a pre-existing deadlock in BenchmarkSelectWithTypeMismatch where defer rows.Close() inside a loop with MaxOpenConns=1 caused connection starvation. Moved rows.Close() to end of loop body.

Testing

• All 96 benchmarks pass locally (go test -run='^$' -bench=. -benchtime=1s . ./msdsn)
• Unit tests unaffected (go test ./msdsn ./internal/... ./integratedauth ./azuread)
• CI workflow validates main-vs-PR comparison end-to-end

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.60%. Comparing base (f92e6f0) to head (0e6b55a).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #376       +/-   ##
===========================================
+ Coverage   80.70%   96.60%   +15.89%     
===========================================
  Files          35       92       +57     
  Lines        6910    74351    +67441     
===========================================
+ Hits         5577    71826    +66249     
- Misses       1064     2190     +1126     
- Partials      269      335       +66     

Flag	Coverage Δ
unittests	96.53% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.
see 59 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.