Optimize morph planning and add precache and fallback diagnostics

[KickNext]

What changed

Preparing a new morph repeatedly recalculated contour metrics and pair costs during assignment. Precompute them once while preserving correspondence order and visual output. Synthetic AOT measurements on Windows show a six-contour plan improving from 441.6 to 65.1 microseconds (median p50 across three runs).

Add precacheMorph for explicit preparation in the existing configurable cache and onFallback diagnostics on both widgets, including release builds. Add CPU/frame-timing harnesses and a reproducible visual A/B capture test. Defaults and animation trajectories remain unchanged.

Verification

• 141 package tests with 91.70% coverage, including cache clearing/disabling/limits, interrupted requests, disposal, fallback and golden tests.
• 27 example tests; analyzer and formatting pass; full release web example builds.
• Chrome smoke test and 13 new API tests pass. Local Flutter test serving returned 404 for CanvasKit; the tests used the identical local SDK assets supplied through CDP without modifying package or SDK code.
• Visual A/B against main 361f54d: 108 paired captures / 3,600 icon frames / 147,456,000 RGBA pixels, zero differences without tolerance. Includes controlled progress, sizes, DPR, RTL, shadows and repeated spring interruptions. Contact sheets also inspected visually.
• Debug Chrome launch and hot reload pass with no runtime errors.
• Publish dry run reports only uncommitted-file warning from the pre-commit run. Local pana 0.23.18 cannot finish because its sandbox rejects Windows drive-letter paths; Ubuntu CI remains the package-score check.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-22T05:00:43.547494Z	81dfcb5	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.