[test] fix flaky testRebalanceWithRemoteLog

[wattt3]

Purpose

Linked issue: close #3385

The test captures the remote log segment count, then asserts it's unchanged after rebalance.
But, produceRecordsAndWaitRemoteLogCopy() does not ensure that all log segments have been copied to remote, so the remoteLogSize(captured count) is non-deterministic.

Brief change log

• Add FlussClusterExtension#waitUntilAllLogSegmentsCopyToRemote(), which waits until the remote log end offset reaches the active segment's base offset (i.e. all non-active segments are copied to remote).
• Uses waitUntilAllLogSegmentsCopyToRemote() instead of waitUntilSomeLogSegmentsCopyToRemote(), so the captured segment count is stable.

Tests

Ran RebalanceManagerITCase.testRebalanceWithRemoteLog 100 times

• Before fix: 9 / 100 failed
• After fix: 0 / 100 failed

[wattt3]

Hi @swuferhong / @wuchong
I saw you've worked on this area before.
So I'd appreciate it if you could take a look.
Thanks!