perf: diagnose iOS snapshot latency jitter

[thymikee]

Summary

Add bounded iOS snapshot timing diagnostics from the XCTest runner into snapshotDiagnostics, including backend, phase, and selected-backend timing.

Keep default snapshot output compact by capping retained samples and phase keys while surfacing larger evidence through debug diagnostics/session artifacts.

Refs #917

Validation

pnpm exec vitest run src/__tests__/snapshot-diagnostics.test.ts src/commands/capture/index.test.ts passed.

pnpm check:quick, pnpm format, pnpm build, and pnpm build:xcuitest passed.

pnpm check:unit was run twice; each full run had one unrelated flaky src/utils/__tests__/daemon-client.test.ts failure, and each failing test passed when rerun in isolation.

Manual iOS simulator evidence on iPhone 17 Pro Max with Settings: 4 Settings-root snapshot -i --json --debug samples and 3 General-subpage samples succeeded, and the session was closed. The first root snapshot was 6486ms end-to-end while runner-reported snapshot work was 248ms; warm root snapshots settled around 121ms end-to-end with runner_total around 115ms, backend.tree around 86ms, and xctest_root_snapshot around 51ms. This points the observed cold jitter at runner/session lifecycle or daemon transport startup rather than accessibility traversal/shaping.

[github-actions]

Size Report

Metric	Base	Current	Diff
JS raw	1.4 MB	1.4 MB	+2.2 kB
JS gzip	445.6 kB	446.2 kB	+621 B
npm tarball	586.2 kB	587.5 kB	+1.4 kB
npm unpacked	2.0 MB	2.0 MB	+5.8 kB

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	28.6 ms	29.9 ms	+1.3 ms
CLI --help	50.9 ms	52.1 ms	+1.2 ms

Top changed chunks:

Chunk	Raw diff	Gzip diff
dist/src/9722.js	+278 B	+80 B