test: stabilize Radix pointer interaction coverage

[neonwatty]

Summary

• wait two presentation frames after the trigger animation finishes so Firefox and WebKit hit-test the settled shadow-DOM trigger geometry
• preserve the real .bd-trigger-label pointer click and existing Radix outside-interaction assertions
• assert the click synchronously enters BugDrop’s open state before waiting for the async form UI

Root cause

The test awaited Animation.finished and clicked immediately. That promise can resolve before Firefox/WebKit present the finished geometry, leaving a narrow compositor/hit-test race. CI artifacts from both failures showed Playwright completed the click but the trigger remained and #title never appeared. Waiting for presentation frames closes that race without increasing timeouts, retrying, or calling BugDrop.open().

Verification

• affected test, 20 repetitions per engine: 60/60 passed across Chromium, Firefox, and WebKit with retries disabled
• full e2e/widget.radix.spec.ts: 15/15 passed across all three engines
• npm run validate: lint, format, both TypeScript configurations, and 2,525 unit tests passed
• npm run knip: passed with existing configuration hints only
• independent reviewer stress: 30/30 plus 50/50 cross-browser executions passed
• code reviewer and test analyzer: no findings; simplifier suggestion to remove the isOpen() assertion was rejected because it directly proves trigger activation and made the prior failure deterministic

Review note

The required Codex review helper was attempted twice with its recursion guard. Both attempts stalled after model-cache/plugin startup warnings and emitted no findings; the second was interrupted after the same condition repeated. Manual parallel review gates and the checks above completed cleanly.

Not merged.