Evidence: refresh terminal PR-249 artifact after checksum changes

Related: #722 and #430.

The required threads-execution-inventory-witness gate is deterministically failing on current main:

• run 33463092639 / job 99717230790
• run 33469127268 / job 99735086043

Both report: checked terminal artifact is stale.

Cause: #722 changed cve/mc-lock-stop-vs-park.js and updated its reference checksum from 909776... to d8ad5e..., but docs/.data/pr249-terminal-execution.json still publishes the old checksum and its source-wide measured_commit remains 709f34b9. Regenerating only the terminal aggregate would combine the new checksum with old per-mode execution evidence, so it would make the gate green without restoring evidence provenance.

Refresh the complete clean ReleaseSafe serialized and parallel-js execution inventories, then regenerate the terminal artifact from the exact measured commit. Verify the terminal check and its release-compatibility consumer. Preserve all denominators and fail-closed checks; do not relabel old timings, hand-edit generated records, or weaken the checksum/staleness gate.

Implemented by 4e9de6e5.

Fresh clean ReleaseSafe evidence measured at 0b6f4d96b4ebc687e4c4121c1e815333cb6af141:

• serialized: 246/246 passed, 0 failed, 0 skipped; 87 cases reached the optimizer
• no-GIL: 253/253 passed, 0 failed, 0 skipped; 92 cases reached the optimizer
• terminal aggregate: 259 executable, 253 promoted, 6 terminal dispositions, 0 blocked
• cve/mc-lock-stop-vs-park.js passed in both modes and now carries checksum d8ad5ed57fd07dc28c6d259e9a867c100f698941db4d204298ff35809ba86b9b

Validated:

• tools/pr249-terminal-execution.ts --check
• tools/optimizer-release-inventory.ts
• zig build docs-link-check (308 files)
• bun run docs:build (50 pages)

The umbrella release-compatibility-check reaches an unrelated pre-existing platform-matrix JSON drift before the PR-249 consumer; its direct PR-249 and dependent optimizer checks both pass. Flips 0 test262 cases.