JobTimeoutTask leak problems

- [ ] 我已经在 [issues](https://github.com/alibaba/arthas/issues) 里搜索，没有重复的issue。

### 环境信息

* Arthas 版本: 3.5.3
* 操作系统版本: xxx
* 目标进程的JVM版本: xxx

### 问题描述
分析Java memory dump时发现有大量未释放的JobTimeoutTask实例：

可能是Job管理存在问题，` jobTimeoutTask.cancel()` 并没有从ExecutorService中取消job
```java
@Override
public Job createJob(InternalCommandManager commandManager, List tokens, Session session, JobListener jobHandler, Term term, ResultDistributor resultDistributor) {
final Job job = super.createJob(commandManager, tokens, session, jobHandler, term, resultDistributor);

/*
* 达到超时时间将会停止job
*/
JobTimeoutTask jobTimeoutTask = new JobTimeoutTask(job);
long jobTimeoutInSecond = getJobTimeoutInSecond();
Date timeoutDate = new Date(System.currentTimeMillis() + (jobTimeoutInSecond * 1000));
ArthasBootstrap.getInstance().getScheduledExecutorService().schedule(jobTimeoutTask, jobTimeoutInSecond, TimeUnit.SECONDS);
jobTimeoutTaskMap.put(job.id(), jobTimeoutTask);
job.setTimeoutDate(timeoutDate);

return job;
}

@Override
public boolean removeJob(int id) {
JobTimeoutTask jobTimeoutTask = jobTimeoutTaskMap.remove(id);
if (jobTimeoutTask != null) {
jobTimeoutTask.cancel();
}
return super.removeJob(id);
}
```

```java
private static class JobTimeoutTask implements Runnable {
private Job job;
...
public void cancel() {
job = null;
}
}
```

Start by tracing the createJob and removeJob entry points shown in the issue, then inspect JobTimeoutTask and the scheduled executor lifecycle. Reproduce the job-removal flow with a memory dump or existing job lifecycle checks. Done means removed jobs no longer retain JobTimeoutTask instances while timeout handling still works.

Written by the indexing model from the issue text.