[Bug] Cluster cache refresh stops permanently after an initial failure

**BUG REPORT** 1. Please describe the issue you observed: - What did you do (the steps to reproduce)? Start the dashboard while the first scheduled cluster info request fails and no cached value exists. - What did you expect to see? The scheduled cache task should retry at the next configured interval. - What did you see instead? `refresh` throws when no old value is available. Because it is passed directly to `scheduleAtFixedRate`, that uncaught exception suppresses every future execution of the periodic task. 2. Please tell us about your environment: - apache/rocketmq-dashboard master - baseline commit `770822bbc812bfd0c2a94aa0128ab6f59b50d076` - JDK 17 3. Other information: Direct callers should retain the current failure signal; only the scheduler boundary should contain transient failures.