Improve app_wait_for stability windows and timeout diagnostics

[shaun0927]

#745 Improve app_wait_for stability windows and timeout diagnostics

Why OpenSafari should reflect this

WebdriverIO and Playwright reduce flakiness with auto-wait semantics and detailed timeout diagnostics. OpenSafari already has app_wait_for, but timeout failures currently return only the query, elapsed time, and poll count. For native and Flutter apps, reviewers need to know whether an element was absent, present-but-hidden, present-but-disabled, unstable during animation, or found under a mismatched app context.

This should be reflected because OpenSafari drives real simulator UI via AX/SimulatorKit. Better wait diagnostics improve stability without changing tap/type behavior.

Scope / how to implement

• Extend app_wait_for with backward-compatible optional diagnostics:
  • stable_ms: require the condition to hold continuously before success.
  • timeout response includes last match count and a small bounded sample of nearest/last candidates with role, label, identifier, visible/enabled, frame, and path.
  • include context metadata already produced by native context classification.
• Preserve existing defaults: condition=exists, timeout 10s, interval 500ms, no stability requirement unless requested.
• Add unit tests around condition/stability evaluation using pure helpers.

Decisions needed before implementation

1. Maximum number of candidate samples in timeout output (suggest 5).
2. Whether stable_ms should apply to not_exists as well as positive conditions.
3. Whether candidate sampling should query broad tree only on timeout or during every poll.
4. Naming: stable_ms vs stability_ms.

Success criteria

• Existing app_wait_for calls remain compatible.
• Optional stability window prevents success until the condition has held for the requested duration.
• Timeout JSON explains why the condition was not met using bounded non-sensitive AX metadata.
• Unit tests cover positive condition, not_exists, visible/enabled failures, and candidate truncation.

Post-merge OpenSafari live validation

• Against a booted simulator app, wait for a visible button with stable_ms=500 and confirm success includes elapsed/polls/stable duration.
• Wait for a nonexistent label and confirm timeout output includes query, polls, context, and bounded candidate evidence.
• Confirm a normal existing app_wait_for call still returns the previous core fields.

Ambiguity review

This issue does not add a new locator system or action retry engine. It only enriches the existing wait tool with opt-in stability and diagnostics.

Direction and necessity review (2026 OSS comparison)

• Aligned: yes — improves existing native/Flutter wait behavior while preserving direct AX/SimulatorKit automation.
• Necessary: yes — wait timeouts are a primary source of flakiness and need actionable evidence.
• Minimal first PR: add stable_ms, bounded timeout samples, and pure unit tests; no locator rewrite or action retry engine.

[shaun0927]

Implemented and merged to develop via PRs #748-#751 as scoped for this issue set. Closing because the requested issue/PR validation workflow is complete on the integration branch.