docs: record data-layer D4 ledger-scaling progress

[StackOverFlow11]

Summary

Docs-only progress sync recording Data Layer D4 (coverage-ledger record_many batch writes + in-process lookup cache) as implemented. No code change.

PR #49 merge status

MERGED into main (mergeCommit 05a86d4). D4 code accepted by Codex; this PR only updates documentation to match.

Changed files (3, docs only)

• docs/data/data_layer_contracts.md — D4 recorded as implemented (CoverageLedger / IntradayCoverageLedger record_many(...) batch writes + in-process lookup cache; parquet paths, column order, and coverage semantics unchanged); section-4 title gains / D4; the unimplemented block moves to D5+ (concurrency / global rate limiter = D5, endpoint schema registry, PanelStore append/partition). D5–D6 stay explicitly unimplemented.
• CLAUDE.md — PR feat(cache): D4 coverage-ledger batch writes + in-process lookup cache #49 added to the merged-PR list; D4 progress bullet; quality gate 643 → 656 passed (+ D4 coverage-ledger-scaling=13); roadmap next data-layer step is D5 concurrency / endpoint schema registry (D4 done).
• AGENTS.md — made byte-identical to CLAUDE.md.

D4 scope (as recorded)

record_many(rows) on both ledgers (one normalize + one atomic write; empty no-op; record() delegates); per-instance frame cache + per-(endpoint, key[, raw_freq]) memo (mtime-invalidated, refreshed on write, read() returns a copy); TushareCache not-ready split writes its 1–2 rows in one record_many. Public methods/paths/columns and coverage semantics (only ok/empty count; failed/not_ready do not) unchanged.

D5 / D6 — NOT implemented

Confirmed. This is a publish + docs-sync task only; no D5 concurrency / thread pools / async / global rate limiter, no endpoint schema registry, no PanelStore append/partition, no data-quality change, no research/backtest work.

Verification (docs PR)

• git diff --name-status main...HEAD → only AGENTS.md, CLAUDE.md, docs/data/data_layer_contracts.md (docs only).
• cmp -s AGENTS.md CLAUDE.md → identical.
• git diff --check main...HEAD → clean; merge-marker scan → NO_MERGE_MARKERS.
• stale-text scan: no 643 passed; no "D4 unimplemented" / "next step D4" residual; 656 passed recorded; D5/D6 not claimed implemented.
• docs-only ⇒ full pytest not required (the D4 suite already 656 passed on feat(cache): D4 coverage-ledger batch writes + in-process lookup cache #49).

PR #38 untouched

Confirmed: #38 OPEN base=feat/i5c-mmp-minute-factor — not merged, closed, retargeted, or edited.