fix(ios): preserve timed pan duration

[thymikee]

Summary

Fixes #1562 by routing single-pointer timed pans through the existing native continuous XCTest drag profile. The dense per-sample planned bridge can report success without delivering pan updates on current iOS simulators.

Two-pointer plans still use the exact synchronized samples, and endpoint-hold fling plans retain their fast-swipe profile. The gesture-plan ADR now documents the platform-specific lowering decision.

Validation

• Regression proof: the focused runner test was red before the fix with XCTAssertEqual failed: ("sampled") is not equal to ("continuousPan"); it passes after the fix.
• Passed focused iOS XCTest: testSinglePointerTimedPanUsesContinuousExecution.
• Passed production iOS runner build: pnpm build:xcuitest:ios.
• Passed pnpm format and pnpm check:affected --run; the affected selector delegated the Swift runner lane to GitHub.
• Verified through the public CLI on an iPhone 17 Pro / iOS 26.2 simulator: a 400 ms timed pan preserved durationMs and executionProfile and moved the Home ScrollView content.

[github-actions]

Size Report

Metric	Base	Current	Diff
JS raw	1.96 MB	1.96 MB	0 B
JS gzip	626.6 kB	626.6 kB	0 B
npm tarball	746.3 kB	746.7 kB	+426 B
npm unpacked	2.61 MB	2.62 MB	+3.2 kB

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	27.8 ms	27.7 ms	-0.1 ms
CLI --help	64.4 ms	65.5 ms	+1.2 ms

Top changed chunks: no changes in the largest emitted chunks.

[thymikee]

Reviewed exact head 8ec77a3d2fd663659f51e95a3ffd89333ea40964 — no actionable findings. The production route correctly limits continuous XCTest synthesis to single-pointer timed pans while preserving endpoint-hold fast swipes and sampled multi-pointer gestures. The amended ADR captures the iOS-specific interpolation tradeoff, the routing regression test is meaningful, CI is green, and GitHub reports a clean merge state. The reported live iPhone run covers the prior false-success symptom. Ready for human review.

Residual evidence caveat: response durationMs/executionProfile describe the plan rather than independently measuring delivered timing; a future validation capture of runner start/end uptime would make that claim directly auditable. No separately authorized cross-vendor review was run.