feat(keywords): expose estimated difficulty evidence

[hubab1]

Summary

• calculate and persist the local estimated-difficulty heuristic during successful ranking refreshes
• keep ranking persistence authoritative when optional estimate validation or revision conflicts occur
• expose fresh, stale, unavailable, and missing states in the keyword table and evidence detail sheet
• add app-scoped MCP summary/detail output, dedicated evidence CSV export, and generation-safe deletion behavior
• keep optional post-commit writes generation-safe across delete and re-add races

Safety and correctness

• estimated difficulty remains clearly labeled as a local heuristic, not Apple Ads or Apple-provided truth
• ranking and refresh statuses remain successful when only the optional estimate is rejected
• stale provider completions cannot attach difficulty or provenance to a replacement track generation
• dedicated CSV output neutralizes spreadsheet formula prefixes in external text
• V4 schema stack was built and signed without launching against the live store

Validation

• 122 focused tests passed
• 404 full-suite tests passed
• signed Debug app build and strict codesign verification passed
• three final independent reviews found no remaining P1 or P2 issues
• app and test targets compile cleanly; the broader standalone MCP dependency-closure debt predates this slice and remains separate

Stack and review

• draft/manual category-C PR
• based on feat(storage): persist estimated difficulty evidence #38 and also requires the estimator core in feat(keywords): add estimated difficulty core #36
• do not merge before the ranking-source, heuristic-semantics, and V4 migration gates are accepted
• no live-store app launch was performed

Attribution

Inspired by the estimated-difficulty concept in WIP PR #8 from @akshaynexus, specifically akshaynexus@5de98ea. This is an independent rewrite with bounded evidence, provenance, transaction isolation, generation safety, UI, MCP, CSV, and tests. Source link only; no co-author trailer.