HDDS-11855. Fix flaky TestContainerBalancerDatanodeNodeLimit#checkIterationResultException

[chihsuan]

What changes were proposed in this pull request?

checkIterationResultException failed intermittently (12/100 in CI) with actual 0L >= 3L. The test asserts that at least 3 container moves fail, but the number of moves the balancer schedules depends on the cluster layout, which TestableCluster builds from an unseeded random generator. On some layouts, fewer than 3 moves are scheduled, so the fixed count of 3 is not something the test can rely on.

Fix: Assert what is true for every layout instead of a fixed number. All moves are mocked to fail, so none can be counted as completed (completed == 0); and whenever a move is scheduled, it must be recorded as either a failure or a timeout. The @Flaky marker is removed.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-11855

How was this patch tested?

• Ran the full TestContainerBalancerDatanodeNodeLimit class locally: 336 tests, 0 failures; checkstyle clean.
• flaky-test-check on the fork, whole class, 100 runs (10×10): all green. https://github.com/chihsuan/ozone/actions/runs/28726374356

[adoroszlai]

Thanks @chihsuan for fixing all known issues in TestContainerBalancerDatanodeNodeLimit.