fix(ios-runner): runMainThreadWork reports MAIN_THREAD_TIMEOUT for work that finished at the timeout boundary #2782

Purpose

In apple/runner/AgentDeviceRunner/AgentDeviceRunnerUITests/RunnerTests+MainThreadWork.swift:94-114, after semaphore.wait times out, the code takes the lock and reads workState.finished. It uses that value only for abandoned-work accounting. It then throws timeoutError() unconditionally at :114. If the main block finishes after the semaphore times out but before the lock is taken, it has already stored its result, and that result is thrown away. A tap that actually happened is reported as MAIN_THREAD_TIMEOUT (see RunnerTests+Transport.swift:287-300). A daemon retry can then perform the gesture a second time.

Line references are at 01328de411.

Required behavior

• On the timeout branch, if workState.finished is true under the lock, return or rethrow the stored result exactly as the non-timeout path does. Do not count the work as abandoned, and do not call onAbandoned.
• Only work that is still unfinished under the lock is abandoned and throws the timeout error.
• Do not change any timeout value.

Completion conditions

• A unit test makes the block finish between the semaphore timeout and the lock. It can do this with an injected wait or a test seam in the existing style. The test asserts that the result is returned and that the abandoned count is unchanged.
• Existing abandoned-work and busy-gate tests still pass.
• Run pnpm check:xctest-selection, build the runner, and run the runner unit tests on the macOS host lane (see docs/agents/testing.md for signing and permissions). Add or change tests under the existing #if AGENT_DEVICE_RUNNER_UNIT_TESTS guard. Do not keep a second list of test names.