[Tiering] Trigger job failover when coordinator epoch changes

### Search before asking

- [x] I searched in the [issues](https://github.com/apache/fluss/issues) and found nothing similar.

### Fluss version

0.9.0 (latest release)

### Please describe the bug 🐞

TieringSourceEnumerator records the coordinator epoch only at startup. After a coordinator failover or upgrade, it continues tiering with the stale epoch. The coordinator rejects subsequent heartbeats, causing tiering tasks to time out and the failuresTotal metric to keep increasing.

### Solution

Validate the coordinator epoch in every heartbeat response. If it differs from the startup epoch, throw an exception to trigger a Flink job failover. The restarted enumerator will register again with the latest coordinator epoch.

### Are you willing to submit a PR?

- [x] I'm willing to submit a PR!

Start by locating TieringSourceEnumerator and the code that processes coordinator heartbeat responses. Verify how the startup epoch is stored and how heartbeat failures are handled. Done means a changed coordinator epoch triggers a Flink job failover, allowing the restarted enumerator to register with the latest epoch, with tests covering the mismatch.

Written by the indexing model from the issue text.