perf(clip): cache rasterized coverage masks

[kivutar]

Summary

• cache the immutable pixel coverage mask for each non-rectangular clip-stack state
• share cached masks across unchanged clones and restore the previous cache on Pop
• reuse the cached mask from Context.applyClipToPaint instead of rasterizing the complete clip for every fill

Motivation

Context.applyClipToPaint rebuilt the entire non-rectangular clip mask on every fill. A downstream fullscreen UI repaint containing a rounded window regressed to about 553 ms at 1.333x scaling after the new clip architecture; reusing the mask brings the same repaint back to about 58 ms.

The cache is state-local, bounded by the live clip stack, immutable after initialization, and synchronized so unchanged clones can rasterize concurrently. Mutations receive a new cache, while Pop restores the previous state and its mask.

Validation

• go test . ./internal/clip -count=1
• go test -race . ./internal/clip -count=1
• go vet ./...
• staticcheck . ./internal/clip
• cached applyClipToPaint: 105-106 ns/op, 16 B/op, 1 alloc/op

[kivutar]

CI is currently failing during action setup because GitHub's codeload endpoints return 429/5xx errors; no failing job reached this branch's code. I retriggered the workflow once by refreshing the head commit, but the same infrastructure outage affected more runners and fork authors cannot rerun upstream workflows. Local validation is green as documented in the PR description.

[kivutar]

Can we re-run the CI on this please?