docs: record data-layer D3 quality checks progress

[StackOverFlow11]

What changed (docs-only, no code)

Records Data Layer D3 (PR #45, merged — mergeCommit 60b3258) as implemented.

• docs/data/data_layer_contracts.md — implemented status D1/D2 → D1/D2/D3; a
  concise D3 paragraph: report-only data/quality/ layer that reports findings
  only and never filters/repairs/mutates or changes cache/feed/backtest
  semantics; data-update integration deferred to D3b. D4–D6 remain
  unimplemented.
• CLAUDE.md — PR feat(data): add report-only data quality checks (D3) #45 added to the merged-PR list; a D3 progress bullet (package;
  daily market/adj_factor checks; 1min intraday checks; deterministic bounded
  renderer; secret-safe redaction fix; D3b deferral; no behavior change); quality
  gate 584 → 625 passed; roadmap D3+ → D3b / D4.
• AGENTS.md — kept byte-identical to CLAUDE.md.

D3 scope summary

A pure, report-only data/quality/ package: report.py (findings model +
deterministic, bounded, secret-safe renderer), market.py (daily market /
adj_factor checks), intraday.py (1min checks), _frames.py (pure helpers). Clean
panels yield zero hard findings; "hard" is a report severity, not a default
exception. No cache / feed / factor / alpha / portfolio / runtime change.

D3b / D4 not started

This is a publish + docs-sync task only. No D3b (data-update quality hook),
no D4 (ledger scaling / batch writes), no schema registry, no concurrency.

Verification (Step 4)

• git diff --name-only main...HEAD → AGENTS.md, CLAUDE.md,
  docs/data/data_layer_contracts.md only (docs-only, no code).
• cmp -s AGENTS.md CLAUDE.md → identical.
• git diff --check main...HEAD → clean; merge-marker scan → 0.
• stale-text scan (D3–D6 尚未实现 / 584 passed) → none (no stale
  unimplemented/gate text).
• Secret scan over the 3 docs: real .config.json token value 0 hits; no
  FAKE_TUSHARE_TOKEN.

PR #38

Left untouched (OPEN base=feat/i5c-mmp-minute-factor) — not merged, closed,
retargeted, or edited.

Docs-only, no code behavior change.