[d-wang-commit]

Description

Parallel SHOW REGIONS queries can observe different stages of an asynchronous region migration. For example, one query can return replicas {1,3} while another returns {1,2,3}. The cluster test result set then throws InconsistentDataException while comparing rows, aborting the migration wait and triggering cluster teardown before its deadline.

Retry this exception in four existing migration waits: Stream consensus-directory preservation, IoTConsensus empty-deletion and compression-ratio checks, and the replica-membership helper shared by tree/table multi-data-directory deletion tests. An inconsistent observation is treated as an unmet condition. The original one-minute or ten-minute deadlines and all final assertions remain unchanged.

Tests

• UT: no new unit tests; existing unit tests were not run because this change only adjusts integration-test polling.
• IT: updated the existing waits and their shared helper; no new test cases. All 7 cases passed on the final commit (0 failures, 0 errors, 0 skipped) with JDK 17 on Windows.
  • IoTDBRegionMigrateNormalForIoTV2StreamIT (3 cases)
  • IoTDBRegionMigrateWithLastEmptyDeletionIT (1 case)
  • IoTDBRegionMigrateWithCompressionRatioIT (1 case)
  • IoTDBRegionMigrateWithDeletionMultiDataDirIT (1 case)
  • IoTDBRegionMigrateWithDeletionMultiDataDirTableIT (1 case)
• Validation uses with-integration-tests,ClusterIT with explicit ClusterIT, TableClusterIT, and DailyIT category selection. Reactor verification for integration-test and its dependencies, including Spotless checks, passed.

Side effects and risks

A persistent metadata inconsistency now fails at the existing polling deadline instead of aborting immediately. Retrying is limited to InconsistentDataException in these migration waits; other exceptions still propagate, and the shared result-set consistency checks remain enabled. Production behavior is unchanged.

---

This PR has:

• been self-reviewed.
• added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
• been tested in a test IoTDB cluster.

---

Key changed/added classes (or packages if there are too many classes) in this PR

• IoTDBRegionMigrateNormalForIoTV2StreamIT: retries transient metadata inconsistencies before checking consensus directories.
• IoTDBRegionMigrateWithLastEmptyDeletionIT: retries inconsistent membership observations during migration.
• IoTDBRegionMigrateWithCompressionRatioIT: retries inconsistent region observations in the post-migration compression-ratio wait.
• RegionMigrateFileAssertions: applies the same retry policy to the shared replica-membership wait.