[Caideyipi]

Extend the daily IT coverage for #18623 on top of feat/pipe-logical-backup-sink.

• Check Tree deletion path/time clipping against a populated target, including preservation of excluded paths and timestamps outside [1,3].
• Check Table deletion with a TAG predicate, other devices, null TAG values, excluded tables, and a rewrite at the same timestamp after deletion.
• Wait for a canary in a committed archive event before issuing realtime deletions; SHOW PIPE reporting RUNNING alone is insufficient for this test barrier.
• Cover restore failure on an existing database, checkpoint recovery after resolving that conflict, and rejection of a checkpoint used with a different target. Retain snapshot, aligned/null data, scope/auth exclusion, dry-run, portable ZIP, and deprecated alias coverage.
• Run only IoTDBPipeLogicalBackupIT in a dedicated PR workflow, with ScalableSingleNodeMode and HighPerformanceMode sources and a ScalableSingleNodeMode target. Upload cluster logs and Failsafe reports on failure.

The Table rewrite assertion currently exposes a cross-stream replay ordering limitation. After DELETE FROM measurements WHERE device='d1' AND time<=4, the source rewrites (time=3, device='d1', reading=300). The source query and the archive both retain that rewrite, but restoring the backup loses it. The single-node diagnostic trace shows:

data-1 #13 DELETE device='d1', time=[1,3]
data-1 #16 INSERT d1@3=300
data-2  #4 DELETE device='d1', time=[1,3]
data-3  #4 DELETE device='d1', time=[1,3]
data-4  #1 DELETE device='d1', time=[1,3]

Restore reads each data stream in stream-ID order, so later streams replay the same broad deletion after the rewrite. Other devices, null TAGs, and out-of-range target data survive. The public logical-backup documentation disclaims a global transaction across streams; this test makes the concrete delete/rewrite consequence visible for product-boundary review. The failing assertion is retained as a diagnostic regression in this test PR. It does not test process crashes.

Validation on Windows, JDK 17:

• Spotless and git diff --check passed; the IT sources compiled.
• Full ScalableSingleNodeMode run: 4 tests, 3 passed, 1 failed on the Table rewrite assertion. Source presence and archived rewrite assertions passed.
• Full HighPerformanceMode run: 4 tests, 3 passed, 1 failed on the same missing rewrite. The focused run of the Table case on 5d2738bca7c also failed with the same replay trace; source presence and archived rewrite assertions passed.
• Surefire tests were explicitly excluded; the class was selected through Failsafe's it.test filter.
• Linux PR CI on commit 5d2738bca7c completed: https://github.com/apache/iotdb/actions/runs/39959079302. Both source configurations ran 4 tests: 3 passed, 1 failed, 0 errors. Both failures reproduce the same archived rewrite followed by deletions from later streams. Cluster logs and Failsafe reports were uploaded for both configurations.