docs(research): C++23 feasibility audit — feature_collector + feature_extractor (Research-0757)

[lusoris]

Summary

• Audits core/src/feature/feature_collector.c and feature_extractor.c for C++23 conversion feasibility, re-evaluating the Wave 7 blockers now that cpp_std=c++23 is global.
• Root cause confirmed: test_feature_collector.c hard-includes feature_collector.c and calls 6 static functions directly (aggregate_vector_{init,append,destroy}, feature_vector_{init,append,destroy}). This blocks a direct .c→.cpp rename.
• feature_extractor.c has no test that includes it directly — straightforward rename candidate.
• C11 atomics in feature_extractor.c are already bridged by the MSVC/gcc guard in feature_extractor.h; no ABI concern on Linux gcc/clang.

Recommended path

1. feature_extractor.c — direct .c→.cpp rename, add extern "C" guards to feature_extractor.h, mechanical cast fixups. 2–3 h.
2. feature_collector.c — extract 6 static functions into feature_collector_internal.h, rewrite test_feature_collector.c to include the internal header instead of the .c file, then rename. 2.5–3 h.

Total effort: 5–7 h. Risk: LOW. No golden-data exposure (no score computation in either file).

Deliverables checklist (ADR-0108)

• Research digest: docs/research/research-0757-feature-collector-extractor-cpp23-feasibility.md
• Decision matrix: n/a — audit only, no implementation decision
• AGENTS.md invariant: no rebase-sensitive invariants
• Reproducer / smoke-test: n/a — audit only
• Changelog fragment: no user-discoverable surface changed
• rebase-notes.md: no rebase impact — doc-only PR

Test plan

• Reviewer reads the digest and confirms the static-fn inventory matches test_feature_collector.c
• Reviewer confirms no other test file includes feature_extractor.c or feature_collector.c directly

🤖 Generated with Claude Code

[lusoris]

Superseded by master merge marathon 2026-05-31.

[lusoris]

Closing stale 05-29 vintage branch (contained conflict markers). Recreating from clean rebase onto current master.