HDDS-15650. Fix snapshotUsedNamespace underflow when FSO directory is deleted and purged

[jojochuang]

What changes were proposed in this pull request?

HDDS-15650. Fix snapshotUsedNamespace underflow when FSO directory is deleted and purged

Please describe your PR in detail:
• The Problem: FSO directories have no blocks, so directory deletion incorrectly evaluated to isKeyNonEmpty = false . This decremented the active
bucket namespace usedNamespace but did not increment snapshotUsedNamespace , causing a mismatch when the directory was later purged (unconditionally
decrementing snapshotUsedNamespace by 1 and underflowing the count to -1 ).
• The Fix: Updated OMKeyDeleteRequestWithFSO.java to shift the namespace quota to the snapshot namespace on deletion if it is either a non-empty key or a
directory (
isKeyNonEmpty || keyStatus.isDirectory() ).
• Test Utils Alignment: Updated mock deleteDir helper in OMRequestTestUtils.java to align with actual deletion quota tracking.
• Regression Test: Added a test case testSnapshotUsedNamespaceAfterDirectoryDeleteAndPurge in TestOMKeyDeleteRequestWithFSO.java to verify that
directory deletion
followed by a purge does not cause negative snapshotUsedNamespace values.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-14435

How was this patch tested?

New unit test

[sadanand48]

Thanks @jojochuang for the finding this , LGTM

[rnblough]

This is a good fix. Adding the isDirectory check is the minimal effective change to switch it to the behavior the current design expects. I appreciate the additional sanity checks on the testing side, in particular the assertTrue >=0 test added to TestOMKeyDeleteRequestWithFSO.java‎ in lines 401-403. We should copy this practice for every counting value Ozone has.

Thanks @jojochuang!