fix(android): return from orientation once the display reports the rotation

[thymikee]

Summary

Android orientation wrote accelerometer_rotation and user_rotation and returned at once, while the display rotated some time later. On the loaded CI emulator that takes seconds, and accessibility reads hang meanwhile.

Evidence: the Android Smoke on #2344, https://github.com/callstack/agent-device/actions/runs/34025424834. wait text landscape right after orientation landscape-left reported wait_capture_stalled with zero readable captures: the helper session request timed out at 5s, the one-shot fallback ran into the wait's deadline, and the failed-step snapshot taken right after was already in landscape (nodes 2400px wide). The rotation had landed; the wait's whole budget went to the transition.

Change

setAndroidOrientation now polls dumpsys display for mCurrentOrientation to match the requested rotation, every 500ms up to 15s, before returning.

• A display that reaches the rotation returns as soon as it does (about 1s on a local emulator with the fixture app foreground).
• A display that never gets there fails the command with the observed rotation and a hint, instead of reporting success. This is a behavior change worth a look: on the launcher, which pins portrait (mCurrentAppOrientation=SCREEN_ORIENTATION_NOSENSOR), orientation landscape-left previously "succeeded" without rotating and now fails after 15s with observedRotation: 0.
• A display that reports no mCurrentOrientation at all is left to the setting, as before.

Validation

• Unit tests (fake adb, sleep mocked, fake clock for the budget): returns once the display reports the rotation after two stale reads; fails with requestedRotation: 1, observedRotation: 0 when it never does; leaves a display that reports nothing to the setting. Package suite 641 tests green; typecheck, oxlint, oxfmt, fallow clean.
• Live on a local Pixel 9 Pro XL emulator with the fixture app foreground: orientation landscape-left 1s, mCurrentOrientation=1; orientation portrait 1s, mCurrentOrientation=0.

What this does not explain

The other Android Smoke failure family (the post-alert canary missing while the device is in landscape although orientation portrait had been confirmed) is still open; #2350 adds the rotation evidence for it. This change removes the transition cost from the command that starts a rotation, which is also where a later flip would be visible if it is a delayed re-application.

[github-actions]

Size Report

Metric	Base	Current	Diff
Installed (including dependencies)	4.50 MB	4.50 MB	+989 B
Package (unpacked)	4.50 MB	4.50 MB	+989 B
Package (download)	1.33 MB	1.33 MB	+251 B

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	27.2 ms	27.2 ms	+0.0 ms
CLI --help	79.6 ms	78.2 ms	-1.3 ms

[thymikee]

The new dumpsys display probe has no timeout derived from the remaining settle budget, so a stuck probe can bypass the intended 15-second loop. Add a hanging-probe regression and script the new ADB call in the provider scenario. Coverage and Integration currently fail on the unscripted shell dumpsys display call.

[thymikee]

Addressed, rebased onto main (one commit):

• Each dumpsys display probe is bounded by what is left of the 15s settle budget (timeoutMs: remainingMs(deadline)), and the sleep between probes is bounded the same way, so a stuck probe ends the settle instead of bypassing it. A probe that times out fails the command with orientation <x> could not confirm the display rotation: <adb error>.
• Regressions: the never-rotates case now also asserts the first probe gets the full 15s and every later probe strictly less; a new hanging-probe case scripts one probe that consumes its whole timeout and throws, and expects one probe, the failure, and the clock at 15s.
• The provider world scripts shell dumpsys display to report the last settings put system user_rotation, so the Android Settings flow rotates the way the settle expects.

Local: platform-android 645 tests, provider Android scenarios 13, root typecheck, oxlint, oxfmt, fallow (against main) clean.