fix(mockmed): stabilize textarea render metrics

[abrichr]

What changed

• make MockMed's #note textarea use an explicit font stack and line height
• refresh the reviewed public-artifact inventory for that shipped CSS asset
• keep the existing strict baseline contract: an unperturbed record/replay must use the template rung with zero heals

Root cause

Exact-main CI run 29845581250 passed 2,818 tests on macOS and failed only TestBaselineControl::test_baseline_passes_all_template. The retained report shows step_008 (the unlabeled Note textarea) falling from template to geometry, producing one governed heal while the workflow still completed correctly.

The failed run's recorded crop was 320x128 at y=226, while the replayed native <textarea rows="6"> was about six pixels taller. MockMed set a font size but left the form control's font family and normal line height to Chromium/macOS UA defaults. Because recording and replay intentionally use separate browser processes, host font metrics could differ and make this no-drift fixture look drifted.

This is fixture nondeterminism, not a relaxation of the resolver or healing contract. The prior green macOS exact-main run used the byte-identical recorded template (sha256: 9c6d01b8...) and resolved all eight anchors by template; Ubuntu on the failed commit also resolved all eight by template.

Impact

The synthetic reference app now has deterministic form-control geometry across the separate record/replay browsers. Product runtime thresholds, healing behavior, and the zero-heal assertion are unchanged.

Verification

• pre-fix local headless repetition: 4/4 passed (confirming intermittency)
• post-fix isolated headless baselines: 4/4 passed, each heal_count=0, template=8
• tests/e2e/test_cosmetic_drift.py: 16/16 passed in 177.12s
• git diff --check

No headed browser, local input injection, production mutation, provider call, or release was used.

[abrichr]

Dependency/integration update for the current exact heads:

• fix(resolver): refuse ambiguous OCR targets #226: 6d0151dc4423aa223a85514ba22199f718d9e3af
• fix(mockmed): stabilize textarea render metrics #227: 7192b0516ddf6563cf7ff922475bf24cd3b56c26

#227's standalone e2e-browser failure exposes the geometry-ambiguity behavior fixed by #226; it is not a reason to replace this CSS stabilization with a fixed-height/low-information crop workaround.

The combined tree passed the focused headless browser suite: 18 passed in 187.84s. Baseline was consistently 8/8 template with zero heals. Under MoveDrift, step_008 resolved at the correct (344, 296) point and the following typed input was verified without retry; the healed rerun returned to 8/8 template with zero heals.

Recommended order: merge #226 first after its own required CI is green, rebase this PR onto the new main, then rerun required/full macOS CI before merging this PR.