[BUG] Client heartbeat scheduler can stop permanently after one full failure

### Current Behavior

Client heartbeat reporting can permanently stop after one heartbeat cycle fails against all configured admin servers.

The heartbeat tasks are scheduled with `scheduleAtFixedRate(...)`:

```java
executor.scheduleAtFixedRate(() -> URIS.forEach(this::sendHeartbeat), 30, 10, TimeUnit.SECONDS);
```

The HTTP register repository throws from the heartbeat path when the last admin server fails or when an access token cannot be resolved:

```java
if (StringUtils.isBlank(accessToken)) {
throw new NullPointerException("accessToken is null");
}
...
if (i == serverList.size()) {
throw new RuntimeException(e);
}
```

For `ScheduledThreadPoolExecutor.scheduleAtFixedRate`, an exception escaping the task suppresses subsequent executions of that periodic task. So a single full-cluster outage or token failure can stop future heartbeats until the client process restarts.

### Expected Behavior

A failed heartbeat cycle should be logged and retried on the next scheduled run. Exceptions should not escape the periodic heartbeat runnable.

### Steps to Reproduce

1. Configure a Shenyu client with HTTP register/heartbeat enabled.
2. Make all configured admin servers temporarily unreachable, or make token acquisition return blank.
3. Let one heartbeat cycle run.
4. Restore admin availability.
5. The scheduled heartbeat task may no longer run because the prior exception escaped the fixed-rate task.

### Code Location

- `shenyu-client/shenyu-client-core/src/main/java/org/apache/shenyu/client/core/disruptor/subcriber/ShenyuClientURIExecutorSubscriber.java`
- `shenyu-register-center/shenyu-register-client-beat/src/main/java/org/apache/shenyu/register/client/beat/HeartbeatListener.java`
- `shenyu-register-center/shenyu-register-client/shenyu-register-client-http/src/main/java/org/apache/shenyu/register/client/http/HttpClientRegisterRepository.java`

Start by reading the scheduling path in shenyu-client/.../ShenyuClientURIExecutorSubscriber.java, then trace heartbeat handling through HeartbeatListener.java and HttpClientRegisterRepository.java. Reproduce a full admin-server failure or blank access token and verify that the failure is logged, the periodic task remains scheduled, and a later run succeeds after availability is restored.

Written by the indexing model from the issue text.