HDDS-16135. Fix remaining intermittent failures in TestKeyLifecycleService

[shuan1026]

What changes were proposed in this pull request?

HDDS-16135 reported intermittent failures in TestKeyLifecycleService$Normal#testNestedFSODirectoryScanResume. After HDDS-16033 / #11145 resolve the most of the issues, that method is stable. The full $Normal class still flakes in two other methods.

testScanStatePiggybackedOnDelete

The test reads OmLifecycleScanState once and asserts scanEndTime != null. After the first scan finishes, the lifecycle policy is still present, the next tick recreates the row with scanEndTime = null (a ~40ms window). The one-shot read can land in that window.

This change waits with GenericTestUtils.waitFor until a completed scan state is visible.

testMultipleDirectoriesMatched

The test renames a directory to refresh mtime and immediately unparks the scan. getSubDirectory() merges the dirTable cache with the RocksDB iterator and does not dedupe live entries, so the same directory can be evaluated twice: the cache copy and the DB copy. The stale copy still looks expired, so a just-renewed directory is deleted and waitFor(deletedDirectory == expected) never settles.

This change calls awaitDirCacheDrained after the rename.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-16135

How was this patch tested?

Before:
TestKeyLifecycleService$Normal#ALL (20x10 ALL) Fail
https://github.com/shuan1026/ozone/actions/runs/33775402686

After:
TestKeyLifecycleService$Normal#ALL (20x10 ALL) Pass
https://github.com/shuan1026/ozone/actions/runs/33849488601

Full CI:
https://github.com/shuan1026/ozone/actions/runs/33849024847

[chihsuan]

Thanks @shuan1026 +1 LGTM