HDDS-15474. Drop redundant TestOmSnapshot linked-bucket subclasses

[chihsuan]

What changes were proposed in this pull request?

This change is related to HDDS-9000, which aims to improve the speed of unit and integration tests.

TestOmSnapshot is one of the slowest tests. It's an abstract suite of ~72 tests. Each concrete subclass re-runs the whole suite on its own MiniOzoneCluster. There are currently 8 such combinations, 4 of which are linked-bucket variants added by HDDS-11705 (#7434) to verify that snapshot operations on a linked bucket act on the source bucket.

That fix works by resolving the bucket link once, up front, at every snapshot entry point, before any layout-specific (isFileSystemOptimized()) or native-lib logic runs. So a linked run exercises exactly the same downstream code as a plain run, plus the same single resolution step. Repeating it for every layout and native setting re-proves the same behaviour and adds no coverage, while costing 3 extra full MiniOzoneCluster suite runs.

This PR deletes the three redundant linked wrappers:

• TestOmSnapshotObjectStoreWithLinkedBuckets
• TestOmSnapshotFsoWithNativeLibWithLinkedBuckets
• TestOmSnapshotWithBucketLinkingLegacy

and keep TestOmSnapshotFsoWithoutNativeLibWithLinkedBuckets as the surviving full linked run (native-off, so it runs both in CI and locally; native-on classes self-skip without the RocksDB native lib). The four plain classes are unchanged.

No test method is removed; the matrix shrinks from 8 to 5 combinations, saving ~12 min of CI wall clock in the snapshot job (measured below).

How coverage is preserved

The link-resolution contract stays covered twice over: the surviving linked class still runs the entire suite through linked buckets, and TestOMSnapshotCreateRequest / TestOMSnapshotDeleteRequest / TestOMSnapshotRenameRequest assert at unit level that preExecute() rewrites the request to the resolved bucket. Per-layout diff output remains covered by the four plain classes.

Tradeoff: OBJECT_STORE and LEGACY no longer run through linked buckets; FSO carries the linked coverage alone. If per-layout linked coverage is preferred, keeping one linked class per layout is a simple alternative.

Per-operation link resolution sites (why linked runs are layout/native independent)

Operation	Resolution site
createSnapshot	OMSnapshotCreateRequest.preExecute()
deleteSnapshot	OMSnapshotDeleteRequest.preExecute()
renameSnapshot	OMSnapshotRenameRequest.preExecute()
listSnapshot	OzoneManager.listSnapshot()
getSnapshotInfo	OzoneManager.getSnapshotInfo()
snapshotDiff / submitSnapshotDiff	OzoneManager.snapshotDiff() / submitSnapshotDiff()
cancelSnapshotDiff	OzoneManager.cancelSnapshotDiff()
listSnapshotDiffJobs	OzoneManager.listSnapshotDiffJobs()
read key via .snapshot/<name>	OmSnapshotManager.getActiveFsMetadataOrSnapshot()

Not touched: the OmSnapshotFileSystem*WithLinkedBuckets classes in SnapshotTests (a separate base, TestOmSnapshotFileSystem). Follow-up to HDDS-10308, but independent of it (disjoint files).

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-15474

How was this patch tested?

Test-only change (deletes three config-wrapper classes, adds nothing).

• hadoop-ozone/integration-test compiles after the deletions; no remaining references to the deleted classes.
• The surviving linked class passes locally after the change: Tests run: 72, Failures: 0, Errors: 0, Skipped: 0 (438s), the same test count as before, confirming linked coverage is intact.
• CI (fork build-branch run): integration (snapshot) passed in 35m46s (2146s) vs 47m41s (2861s) for the same job on the same day's master run, ~12 min (~25%) faster. All 5 surviving classes green with the expected skip pattern (12 layout-guarded skips on OBJECT_STORE/LEGACY, 0 on FSO); the TestOmSnapshot matrix dropped from ~2046.8s (8 classes) to 1309.7s (5 classes).

[adoroszlai]

@sadanand48 @smengcl would you like to take a look?

[adoroszlai]

Thanks @chihsuan for the patch.