[flink] Retry Kafka topic cleanup timeouts in tests

[dwangatt]

Purpose

Extracted from #9370, which is being split into smaller, independently reviewable PRs.

This PR contains only the Kafka test cleanup reliability improvement. It is independent of the per-partition bucket-count feature and is submitted separately to reduce the scope of the original PR.

Changes

• Retry Kafka topic deletion when cleanup temporarily times out.
• Add unit coverage for retry behavior in KafkaActionITCaseBaseTest.

Testing

• Existing Kafka CDC integration tests will cover the cleanup path in CI.

Related

• Extracted from [WIP] Split partition-level bucket-count support into focused PRs #9370

[JingsongLi]

Closing after review. This PR changes only Kafka integration-test teardown. The added unit tests prove that the retry helper retries a mocked timeout, but the PR does not link a failing CI run or reproduce a Kafka cleanup failure that this change resolves. I cannot establish end-to-end CI value for this standalone change. Please reopen with a failing run or an integration regression that demonstrates the cleanup benefit.