Record the preregistration merge commit in the handoff - #52

The preregistration merged as c50aa93b95e6fe4a6aa6251491d3c2b5a9480ebe. This records that public freeze and corrects the release handoff before implementation begins.

Evidence boundary

• both declared checkpoint lineages failed the necessary retained-JSONNav-utility condition at 0/64;
• no starting checkpoint was selected;
• Granite Guardian was used only as an unqualified diagnostic; its qualification was not run;
• PairRM qualification, teacher qualification, continuation training and method-level preference evaluation were not run;
• the reserved final test was not accessed.

The release handoff now requires three explicit evidence corrections: preserve and correct the fallback lineage label with provenance, disclose that the separately generated selection manifests are task-identical, and keep unqualified judge values out of headline claims.

Issue #39 will receive an evidence comment after release and remain open because method-level and final-test acceptance gates were not executed.

No threshold, quantitative result, endpoint, selection decision or artifact changed in this PR.

Validation

• ruff check .
• ruff format --check .
• mypy src/miniverl
• python scripts/check_text_integrity.py
• python scripts/release_state.py --check
• python scripts/check_markdown_links.py
• mkdocs build --strict
• pytest -q -m "not gpu and not network" --cov=miniverl --cov-fail-under=80: 2083 passed, 6 skipped, 21 deselected, 86.29% coverage