[fluss-server] Fix replicasOnOffline deadlock in addFetcherForReplicas

### 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 🐞

When a tablet server receives a `NotifyLeaderAndIsr` request for a bucket with no elected leader (`leaderId == null`), the original code threw a `STORAGE_EXCEPTION`. The coordinator then marked that replica as offline via `replicasOnOffline`. Since `isReplicaOnline()` excludes replicas in the offline set, subsequent elections could never select these replicas as leader — causing a permanent no-leader state.

### Solution

Replace the error branch with a guard condition (`leaderId != null && leaderId >= 0`). When no valid leader exists, skip fetcher setup silently and let the next `LeaderAndIsr` notification recover naturally.

### Are you willing to submit a PR?

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

Start at the addFetcherForReplicas entry point in the Fluss server code and trace how NotifyLeaderAndIsr handles a bucket with no elected leader. Replace the error path with the stated validity guard so replicas are not marked offline and fetcher setup is skipped for an invalid leader. Confirm that a later LeaderAndIsr notification can recover the replica without a permanent no-leader state.

Written by the indexing model from the issue text.