集群流控下，如何保障TokenServer高可用性

## Issue Description
无论是采用alone还是embedded，当一台TokenServer挂了，是否有机制自动选择其他可用的Server。例如embedded时，如果机器A作为TokenServer挂了，必须要手动调用http://:/setClusterMode?mode=指定一台新的TokenServer吗，是否有自动选举功能？如果没有，是否有告警通知
Type: *bug report* or *feature request*

### Describe what happened (or what feature you want)

### Describe what you expected to happen

### How to reproduce it (as minimally and precisely as possible)

1.
2.
3.

### Tell us your environment

### Anything else we need to know?

Start by reading the cluster-mode handling around the /setClusterMode?mode= entry point and the TokenServer implementations for alone and embedded modes. Determine whether failover, leader election, or alerting is already supported; done would require a documented and testable high-availability behavior for a failed TokenServer.

Written by the indexing model from the issue text.