[BUG] Cluster master election scheduler can stop after one renewal exception

### Current Behavior

Cluster master election/renewal can stop permanently after one exception escapes the scheduled task.

`ShenyuClusterService.startSelectMasterTask()` schedules master selection with `scheduleAtFixedRate(...)`:

```java
executorService.scheduleAtFixedRate(() -> doSelectMaster(host, port, contextPath), ...);
```

`doSelectMaster()` catches exceptions, closes the check services, and then throws a new `ShenyuException`:

```java
} catch (Exception e) {
LOG.error("select master error", e);
upstreamCheckService.close();
instanceCheckService.close();
throw new ShenyuException(...);
}
```

For `ScheduledThreadPoolExecutor`, an exception escaping a fixed-rate task suppresses later executions of that periodic task. A transient lock/renewal failure can therefore stop future master selection attempts in that process.

### Expected Behavior

Master election/renewal errors should be logged and the next scheduled election attempt should still run. Exceptions should not escape the periodic scheduler task.

### Impact

After a transient cluster backend failure, the admin node may stop participating in master election until it is restarted.

### Code Location

- `shenyu-admin/src/main/java/org/apache/shenyu/admin/mode/cluster/service/ShenyuClusterService.java`
- `startSelectMasterTask()` schedules the periodic task.
- `doSelectMaster()` rethrows from the catch block.

Start by reading shenyu-admin/src/main/java/org/apache/shenyu/admin/mode/cluster/service/ShenyuClusterService.java, focusing on startSelectMasterTask() and doSelectMaster(). Trace how the scheduled task handles renewal exceptions; done means errors are logged without terminating future scheduled election attempts.

Written by the indexing model from the issue text.