[Test] Fix deletion regression test flush control

[Caideyipi]

Description

The deletion regression tests set managedByFlushManager before calling asyncClose(), so its guard returns before queuing the working memtable. This breaks both deletion WAL tests and causes a null-future exception during cleanup of the flush-completion test.

Claim manual flush control in the flush-time callback, after the close guard and before FlushManager registration. This lets the tests exercise the production close path while retaining control over flush execution and checking deletion persistence.

Validation

• Previous validation of this fix: TsFileProcessorTest — 26 tests passed.
• mvn spotless:apply -pl iotdb-core/datanode — passed.
• git diff --check origin/master...HEAD — passed.

---

This PR has:

• been self-reviewed.
• added comments explaining the intent of the flush-control callback.
• modified existing unit tests to exercise the close path and verify deletion WAL records and persisted modifications.

---

Key changed classes

• TsFileProcessorTest

[hongzhi-gao]

Reviewed the close/flush callback sequence and the assertions in the affected deletion regression tests. No blocking code-review findings. My local targeted run did not reach tests because of stale cross-module build artifacts; DataNode unit-test CI is still pending, so please wait for it before merging.