fix(rate-of-closure): harden forgiveness evidence contracts

[dieterolson]

Summary

• preserve unsupported turf outcomes and apply explicit out-of-domain penalties/constraints
• convert ordinary per-trial retained post-processing faults into auditable failure records while preserving cancellation
• strengthen browser turf DbC and reject nonfinite evidence in statistics and JSON/CSV exports
• prevent custom-target objective ID aliasing by deriving stable IDs from the actual target at nine-decimal precision
• document the retained-run evidence and failure boundaries

Verification

• python -m pytest tests/rate_of_closure/test_chip_forgiveness_analysis.py tests/rate_of_closure/test_chip_forgiveness_runner.py tests/rate_of_closure/test_variation_gui.py -q (29 passed)
• focused Ruff check and format check
• focused browser tests (3 files, 19 passed)
• npm run type-check
• npm run lint -- --max-warnings=0
• git diff --check

Stack and remaining scope

• base: feat(rate-of-closure): add chip forgiveness decision analysis #4190 chip forgiveness decision analysis
• this is a bounded robustness repair; multi-wedge Pareto campaigns, expanded sensitivity/dispersion evidence, browser worker execution, performance budgets, cross-client bootstrap identity, and full UI parity remain open under Wedge analysis: Robust chip-shot variability and forgiveness study #4164

Refs #4164

[github-actions]

⚠️ SPEC.md Update Required

Source files were modified in this PR but SPEC.md was not updated.

Per the Specification-Driven Development Policy:

• Any PR that adds, removes, or changes functionality must update SPEC.md
• Update the relevant sections (features, architecture, tests, dependencies, etc.)
• Bump the Spec Version if making substantive changes

To resolve:

1. Update SPEC.md to reflect your changes, OR
2. Add the spec-exempt label if this PR genuinely doesn't affect the spec (e.g., pure refactor with no behavior change)

This check is enforced by the spec-check CI workflow.

[dieterolson]

Superseded by #4466, which consolidates all 43 open src/rate_of_closure PRs into one
branch. Closing this one is measured, not assumed: verify_coverage.sh reports
safe=43 NOT-covered=0 against consolidated/rate-closure-remainder-2026-08-13, so this
PR's head is contained and nothing here is dropped.

Stacks were folded top-down (each child into its own parent, then the parent upward), so
non-tip work is included rather than skipped — including 8067ca79a (#4171, CAD export
artifact validation) and f9afa9d96 (#4309, strict ground result playback), each of which
lived in exactly one PR and is absent from its stack tip.

The consolidation also repairs defects the merge exposed, listed in #4466: two unguarded
datetime.UTC imports that break the Python 3.10 lane, a PyQt module registry that could no
longer construct the main window, the club-assembly binding and STL/sidecar controls that had
been dropped from the wired tree, and a filelock dependency that would have failed
tests (3.11) at collection.

Reopen this PR if you need its branch; the branch itself is untouched.