Stabilize flaky CleanupManagerTest.testCleanupHappens

[vharseko]

Problem

CleanupManagerTest.testCleanupHappens is flaky. It offers 500 cleanup actions and then waits for them to run — but the wait loops only while the queue is non-empty:

cm().setPollInterval(100);
for (int i = 0; i < 10 && cm().getEnqueuedCount() > 0; i++) {
    Thread.sleep(1000);
}
assertEquals(500, _counter);

getEnqueuedCount() drops to 0 as soon as the background CLEANUP_MANAGER thread dequeues the batch — which happens before performAction runs. So on a slow/loaded runner the wait loop sees getEnqueuedCount() == 0 on the first check, exits without sleeping, and asserts while no action has been performed yet:

java.lang.AssertionError: expected:<500> but was:<0>
    at com.persistit.CleanupManagerTest.testCleanupHappens(...)

Observed on build-maven (ubuntu-latest, 17); passes elsewhere.

Fix

Wait on the manager's own performed/error counters (getPerformedCount() + getErrorCount()), which advance only after each action completes, until all 500 are processed — with a 30 s cap. This waits for the actions to actually run rather than merely leave the queue. All four existing assertions (_counter == 500, errorCount == 1, performedCount == 499, in-order) are unchanged.

Not caused by this PR's parent change

Unrelated to the PR it was blocking (#228, make Key.SDF thread-safe): that change is a SimpleDateFormat → ThreadLocal in Key, nothing to do with the cleanup manager. CleanupManagerTest was previously touched by #240 for a different race (out-of-order recording), not this queue-drain wait.

Verification

• testCleanupHappens 6/6 green; full CleanupManagerTest class 3/3.
• Root cause precisely identified (queue-count drops at dequeue, before performAction), so this is a deterministic fix.