compare's migrated cross-source checks drop --since's changed-path confidence boost

Index: Known gaps

Status (re-verified 2026-10-09 at 456989f): OPEN. workflows/cross_source_hygiene.py:360 is still def compute_cross_source_evolution(old, new) with no changed_paths parameter, and checker.py has no changed_paths.

Found during the ADR-068 Phase 6 doc follow-up (a stale mention of the retired scan --since in catalog/cases/case181_xcheck_public_to_internal_dependency/README.md prompted checking whether the claim still held).

buildsource/cross_source_checks.py's CrosscheckConfig.changed_paths lets a check (currently public_to_internal_dependency) report a finding at higher confidence when the internal declaration it reaches lives in a file the revision actually changed — "this call reaches a file that changed this revision" is a stronger signal than "this call reaches something internal". The retired scan --since wired its changed-path set through to this automatically.

compare()'s own migrated cross-source-evolution path (checker.compare() → workflows.cross_source_hygiene. compute_cross_source_evolution(old, new)) does not currently accept or forward a changed-path set at all — the function's signature is (old: AbiSnapshot, new: AbiSnapshot), with no changed_paths parameter, even though compare's own --since/--changed-path flags exist and are already threaded through to other consumers (cli_compare_helpers.py's changed_paths parameter, ADR-043 D7 POI scoping — comment-tagged # ADR-068 Phase 2c: ADR-043 D7 POI scoping). So public_to_internal_dependency is always reported at the lower "reaches something internal" confidence under compare today, regardless of --since/--changed-path.

Tractable when picked up, but checker.compare() has no changed_paths parameter to thread today (verified against its real signature, abicheck/checker.py's def compare(old, new, suppression=None, *, policy=..., ...) — no changed_paths anywhere in it) — an earlier version of this entry wrongly described the value as already existing on compare() and needing only to be threaded further down. Adding a changed_paths parameter to checker.compare() (and updating every caller that already resolves a changed-path set — cli_compare_helpers.py already has one — to actually pass it through) is itself the first piece of required wiring work, not a step that can skip straight to modifying compute_cross_source_evolution/CrosscheckConfig. Needs a regression test asserting the confidence does change under --since/ --changed-path on a real (not just internal-API) compare invocation, not only an internal run_crosschecks(...) call — the gap here was invisible to internal tests precisely because nothing exercises the public --since flag's effect on this specific check's confidence.