Add helper-level rev5 vs rev6 validator, correctness-first monte_carlo_super_bs_eirp_dist_rev7, and rev15 integration

[nicklasorte]

Motivation

• rev14 introduced a faster helper (rev6) that produced unacceptable output drift versus rev11 by replacing per-row interp1('spline') with custom PP evaluation, so a correctness-first path and diagnostics are required.
• Provide a deterministic harness to compare rev5 and rev6 helper behavior on identical inputs to locate semantic divergence points before any further optimization.
• Restore a safe production path by wiring a rev7 helper that preserves rev5 semantics and integrating it into the subchunk flow as rev15 for end-to-end validation.

Description

• Added validate_monte_carlo_super_bs_eirp_dist_rev5_rev6.m, a helper-level comparator that reconstructs the exact helper-call inputs, runs both rev5 and rev6, reports shapes, max/mean absolute and relative diffs, worst-case indices, and clustering near endpoints/breakpoints/clamps, and fails closed on tight thresholds.
• Implemented monte_carlo_super_bs_eirp_dist_rev7.m as a correctness-first helper that preserves rev5 per-row interp1(...,'spline') semantics, adds shape/length assertions and optional DEBUG_CHECKS, clamps queries, and avoids the manual PP evaluation strategy used in rev6.
• Implemented subchunk_agg_check_maxazi_rev15.m by cloning the rev11 aggregation/RNG/chunking flow and replacing only the EIRP helper invocation with monte_carlo_super_bs_eirp_dist_rev7, preserving RNG and chunking behavior.
• Added validate_subchunk_agg_check_maxazi_rev11_rev15_statistical.m to run end-to-end rev11 vs rev15 comparisons on identical inputs, measure runtimes, compute summary and upper-tail metrics (mean/std/min/max/median/p90/p95/p99), and enforce fail-closed thresholds.

Testing

• Basic repository checks and file inspections were performed and confirm the four new files are present and syntactically viewable in the tree. (creation and content printed for review). — succeeded.
• Static consistency: the new validators use must_exist(...) guards and return structured results objects for automated gating; those functions are in place for runtime validation. — code-paths added (no runtime execution).
• MATLAB/Octave runtime validation could not be executed in this environment because octave/MATLAB is not available, so helper-level and end-to-end numeric tests were not run here; the validation scripts are written to be run in a MATLAB/Octave environment and will fail-closed if drift thresholds are exceeded. — runtime check not executed (environment limitation).

---

Codex Task