[Lake/Iceberg] Fix testLogTableCompaction flakiness by adding compaction timeout and retry logic

[hemanthsavasere]

Linked issue: close #2867

The testLogTableCompaction test in IcebergRewriteITCase was flaky due to two root causes:

1. Indefinite blocking: IcebergLakeWriter.complete() called compactionFuture.get() without a timeout, which could block indefinitely if the async compaction was slow or stuck.
2. Tight assertions: After triggering a write+compaction, the test immediately called checkFileStatusInIcebergTable() without any retry, causing it to fail if the compaction commit had not yet been applied.

Brief change log

• IcebergLakeWriter: Added a 5-minute timeout (COMPACTION_TIMEOUT_SECONDS = 300) to compactionFuture.get(). On timeout, logs a warning and cancels the future rather than hanging indefinitely.
• IcebergRewriteITCase: Updated testLogTableCompaction to use a 2-minute assertReplicaStatus timeout (via a new overload) when waiting for the write+compaction round to complete, and switched to waitForFileStatusInIcebergTable() which retries the file-count assertion.
• FlinkIcebergTieringTestBase: Added two helper methods:
  • assertReplicaStatus(TableBucket, long, Duration) — timeout-aware overload of the existing method.
  • waitForFileStatusInIcebergTable(TablePath, int, boolean) — wraps checkFileStatusInIcebergTable in a 2-minute retry loop.

Tests

• IcebergRewriteITCase#testLogTableCompaction — the flaky test itself; now uses retry-based assertions and a longer timeout to tolerate compaction latency.

API and Format

No API or storage format changes.

Documentation

No new features introduced; this is a stability fix. The Helm docs change is cosmetic table reformatting only.