Fix TempLocationManagerTest flakiness

[jbachorik]

What Does This Do

Stops TempLocationManagerTest.testConcurrentCleanup from depending on the shared AgentTaskScheduler. Cleanup now runs on a dedicated thread owned by the test, with a try/finally that always drains the worker on assertion failure.

Also tightens the test suite:

• testCleanup: replace the buggy assertFalse(rslt && Files.exists(...)) (which silently passed when cleanup failed) with explicit assertTrue(rslt) + assertFalse(Files.exists(...)). Replace the fragile walk-up loop with TempLocationManager.getBaseTempDirName().
• testCleanup: route through the existing private instance(...) helper for consistency.
• testCleanupWithTimeout: drop the unreachable visitFileFailed override (no @MethodSource argument ever selects it).
• testConcurrentCleanup: hooks now return the CleanupHook default rather than null, register baseDir.deleteOnExit(), and add an assertFalse(cleanupThread.isAlive()) sanity check.

Motivation

Recurring CI flake: Cleanup thread should reach hook within 30 seconds expected: <true> but was: <false>. Root cause is the test's cleanup task being queued in the shared AgentTaskScheduler worker behind cleanup tasks scheduled by other tests in the same JVM.

Additional Notes

Verified by running testConcurrentCleanup 5 times in a row locally — all green.

Contributor Checklist

• Format the title according to the contribution guidelines
• Assign the type: and (comp: or inst:) labels in addition to any other useful labels
• Avoid using close, fix, or any linking keywords when referencing an issue
• Update the CODEOWNERS file on source file addition, migration, or deletion
• Update public documentation with any new configuration flags or behaviors