[FLUSS-3716] Tolerate removed tablet servers in wait helpers

[wolfkill]

What changed

FlussClusterExtension now treats tablet servers that are still referenced by LeaderAndIsr but have already been removed from the local test cluster as a retryable wait condition instead of throwing a NullPointerException.

This covers the remaining wait-helper call sites from FLUSS-3716:

• waitUntilTableReady(...) via waitReplicaInAssignmentReady(...)
• waitUntilTablePartitionReady(...) via waitReplicaInAssignmentReady(...)
• waitUntilAllReplicaReady(...)

Why

During tablet-server shutdown or failover, ZooKeeper metadata can temporarily still reference the removed server while the in-process FlussClusterExtension map has already dropped it. CommonTestUtils.retry(...) retries AssertionError, but the old direct getTabletServerById(...).getReplicaManager() call produced an immediate NPE, turning a transient state into a flaky test failure.

Validation

• RED: JAVA_HOME=/Users/cc/Library/Java/JavaVirtualMachines/corretto-18.0.2/Contents/Home mvn -pl fluss-server -am -Dtest=FlussClusterExtensionTest#testWaitUntilAllReplicaReadyRetriesWhenIsrContainsRemovedTabletServer -DfailIfNoTests=false test
  • failed before the fix with NullPointerException from FlussClusterExtension.waitUntilAllReplicaReady(...)
• GREEN: same targeted test passed after the fix
• JAVA_HOME=/Users/cc/Library/Java/JavaVirtualMachines/corretto-18.0.2/Contents/Home mvn -pl fluss-server -am -Dtest=FlussClusterExtensionTest,StopReplicaITCase,TabletServerShutdownITCase -DfailIfNoTests=false test
  • Tests run: 10, Failures: 0, Errors: 0, Skipped: 0
• JAVA_HOME=/Users/cc/Library/Java/JavaVirtualMachines/corretto-18.0.2/Contents/Home mvn -pl fluss-server -am -DfailIfNoTests=false test
  • fluss-server: Tests run: 1135, Failures: 0, Errors: 0, Skipped: 0
  • reactor result: BUILD SUCCESS
• git diff --check