batch cs shutdown

- [x] mark the server as Readonly, prevent write/recover to this server
- [x] kick the server on shutdown list one by one. When hi_recover list is empty, then we can kick the next one
- [ ] make sure no hi check appears during the shutdown process(may caused by server down)
- [ ] Add an web interface for user
- [x] 将待下线机器标记为Readonly，保证后续不会有新的写流量及副本恢复的block push到该机器上
- [x] 按下线列表对chunk server进行kick，待hi清零后，kick新机器
- [ ] 保证在下线过程中，不出现hi check
- [ ] 使用户在web页面上可以随意下机器

Start by tracing the batch CS shutdown flow, including the shutdown list and hi_recover handling. The remaining work is to prevent hi checks during shutdown and provide a web interface for users to select machines for shutdown.

Written by the indexing model from the issue text.