Update CTC-01 slow guards for member-specific PMI

[pzfreo]

Follow-up to #2174. The post-merge slow lane found two CTC-01 tests whose assumptions changed with member-specific STEP diameter PMI.

What changed

• The Sheet script test now checks that five source diameter tolerances attach to their pattern members, including eight member-specific bore references. It still checks the independent hole tolerances and the remaining angular measured dimension.
• The dimension-lane canary now obtains declaration identities from the generated script's named features and source IDs before applying overrides. This keeps the test attached to the same physical features as declaration numbering changes.
• Its replay subprocess imports this checkout's source tree. With the correct control-frame targets, the authored lanes preserve requirements, completeness, and fidelity and improve legibility for one feature-leader crossing; the assertion follows that measured result.

Evidence

• Exact pre-merge 09bafdf5: the old Sheet test passed. Exact merged ee001d57: it failed because five measured-dimension calls became pattern-member requirements.
• Main post-merge run 37170029864 failed only these two slow tests; other slow jobs and required PR checks passed.
• Both corrected slow tests pass individually on the merged implementation. scripts/pr-check --static passes on the final test-only branch.
• Fresh independent exact-head review of 658ed1f9 found no actionable issue. Required PR CI must pass before merge.

Architecture

No runtime behavior or ADR invariant changes. The tests continue to check ADR 4 replay meaning and ADR 5 source provenance, while the lane canary checks ADR 2 placement behavior through the generated Sheet script.