P3: provenance, release and real-data figures (REP-02, REP-03, S-RP.1-4, FIG-02, SCI-07) #41

Package P3 of the iteration-3 plan: provenance, release and the real-data figures. Owner: agent for code and figures; blocked in part by #46. Size: days.

Closes or advances: REP-02, REP-03, S-RP.1 to S-RP.4, FIG-02, SCI-07, COMP-01 (archive part), S-RE.5/S-FD.4/S-FD.5.

• Light CI (build and sidecar checks). A workflow that builds the PDF and fails on ??, ,%, << and [Funding in its text, runs the fast figure generators and diffs their arrays against the committed sidecars (add a --check mode to codameter.figures), and runs the test suite with the pinned environment. The reviewer itself stays on demand.
• Sidecar provenance (S-RP.1). figures.py: record a dirty-tree flag (git status --porcelain) and a generator source digest (as golden._generator_hash does) beside git_commit; rebuild every figure from a clean tree at the release tag.
• Shard provenance (S-RP.4). bench.py score_cell rows carry golden._generator_hash() and the commit; check_shards requires agreement.
• Calibration invocation (S-RP.3). Data availability names the three producing invocations (--n 200 --start-seed 2000, scenarios clean, shared_source, clock_drift, ~58 min each with single-threaded BLAS) or points to the settings and seeds fields of each locked_*_n200.json.
• Field products (REP-02). Once the author supplies rights and target: archive the 2-4 Hz daily products with sha256 and a licence file; cite SCEDC and the Clements and Denolle 2023 data release with DOIs; pin the noisepy-dvv-cloud commit and --use-case in paper/data/gate1/README.md.
• Comparison figure (S-RE.5, S-FD.4, S-FD.5). Write an in-repo generator for Fig 15 from the archived products and paper/data/gate1/comparison.json with the corrected error bars and the matched-rule r; register it in codameter.figures; keep the interferogram and warm-up figures external and say why (CCFs not archived); fix qmd 1299-1300 ("near 0.7") against Table 6.
• Release (REP-03, S-RP.2). Bump the version (the Weaver floor and scorer changes are breaking; the docstring already says "Before v0.5"); tag from a clean tree after regenerating every figure and the calibration table at that tag; deposit through release.yml (Zenodo); cite tag and DOI in Data availability; update CITATION.cff (currently 0.1.0 with a JGR preferred citation) and the README zenodo placeholders.

Acceptance: a reader can regenerate every generated figure, the calibration table and the comparison table from the tagged release; the only external artefacts are the two interferogram-based figures, named as such.

Working agreement (agents and humans). One PR per issue, branch iter3/<package> off docs/sign-convention-manuscript; name the ledger IDs in the PR title; request Copilot review before approval. Do not edit review/REVISION_PLAN_ITER3.md except to tick boxes and append a dated log line naming this issue. Do not hand-edit statuses in reviews/codameter-gji.review.json; the reviewer reconciles them in P7. Tests and pre-commit hooks must pass. Before requesting review, rebuild the PDF (cd paper && ../.pixi/envs/dev/bin/python build.py --no-survey) and check its text for ??, ,%, << and [Funding. Evidence for every finding: reviews/iter2/block_S-*.md; full report reviews/codameter-gji.iter2.report.md; plan review/REVISION_PLAN_ITER3.md.