fix(preflight): report isolated over-limit ink regions (#116)

[mberrys]

Summary

Closes #116 (L01-11). ink-coverage no longer suppresses over-limit regions below a percentage of the page. The floor is now physical: min_region_area_mm2 (default 0.25 mm², about a 0.5 mm square, the smallest solid a press reproduces as a distinct area). A probe_dpi too coarse to resolve that floor (fewer than four raster pixels) reports the page check-incomplete instead of passing clean. min_region_area_pct is deprecated, still accepted, and ignored.

Proof

Source SHA under test: head of cc/issue-116-ink-coverage-floor. Fixture loop-preflight/testdata/fixtures/ink-coverage-isolated-region.pdf (generator loop-preflight/tools/generate_ink_isolated_region_fixture.py) is one 6 mm square of 400% rich black on an A3 page (about 0.03% of the page), committed first (126daa5).

	Result (PdfTool preflight, test-ink-coverage.json)
Before	pass: true, 0 errors, 0 warnings
After	1 ink-coverage warning: 38.2 mm², peak 400%

The baseline PdfTool was built from the #157 head, which does not touch ink-coverage code.

Local (Windows, MSVC Release): UnitTestsPreflightEngine, UnitTestsPreflightCorpus, UnitTestsPreflightVerdict, UnitTestsPreflightChecks, UnitTestsEvidenceGraph, UnitTestsContentProcessorLimits all exit 0. Regenerating the corpus snapshots changed only the new ink-coverage-isolated-region.json (other differences were line endings). generate-architecture-catalogs.py --check, check_source_integrity.py, check_preflight_truth_source.py and check-change.py --dry-run pass. New tests cover the isolated region on a large page, the deprecated percentage floor no longer suppressing it, an unresolvable floor reporting incomplete, and min_region_area_mm2 validation.

Not run locally

Full check-change build/ctest/clang-tidy (no clang-tidy-18), Linux lanes, and the remaining mapped test targets; left to CI.

Remaining risk

• Regions below the floor stay suppressed by design; the floor is a policy choice.
• The probe is a raster estimate, not a plate-ready separation TAC.
• A profile that relied on min_region_area_pct to hide small regions will now see them.

Handoff

Changed: probe and engine (pdfinkcoverageprobe.*, preflightengine.*), schema, example and test profiles, README, overlay (ink-coverage-raster-tac landed, closed_by ink-coverage), regenerated catalogs, fixture, generator, snapshot, change record and evidence. Anti-slop review: the physical floor replaces the percentage rather than adding a second knob, and the incomplete path fires only when the raster cannot resolve the configured floor.

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[mberrys]

Incomplete finding is emitted on every page, even when nothing is over the limit. The schema says an over-limit region the raster cannot resolve makes the check incomplete. This branch instead fires whenever the floor is below the raster resolution, even with overLimitAreaMM2 == 0. With min_region_area_mm2: 0 (allowed by the schema) every page of a clean document is reported incomplete. Gate it on the page actually having over-limit pixels.