[Bug]: 在切换模型时执行 "zero-downtime reload"，会重启 agent，但钉钉 stream 线程可能没有被正确停止

## CoPaw Version

v1.0.0，v0.2.0
[Using `copaw --version` in your command line or checking the version in the console UI can help.]

## Description

[Describe the bug clearly and concisely. What happened vs what you expected?]
正常情况copaw占用cpu很低，在百分之零点几的量级，在聊天界面切换模型后进程会占用百分之十几的CPU，也就是基本占满单核，更换模型后，旧的钉钉 stream 线程没有正确停止！

**Related PR(s):** #(optional)

**Security considerations:** [If applicable, e.g. auth, env/config exposure]

## Component(s) Affected

- [ ] Core / Backend (app, agents, config, providers, utils, local_models)
- [ ] Console (frontend web UI)
- [ ] Channels (DingTalk, Feishu, QQ, Discord, iMessage, etc.)
- [ ] Skills
- [ ] CLI
- [ ] Documentation (website)
- [ ] Tests
- [ ] CI/CD
- [ ] Scripts / Deploy

## Environment

- **CoPaw version:** [e.g. 0.x.x or git commit]
- **OS:** [Windows 10]
- **Install method:** [ one-line install ]
- **Python version (if applicable):

## Steps to Reproduce

1.
2.
3.

## Actual vs Expected

- **Actual:**
- **Expected:**

## Logs / Screenshots

[Paste relevant log output or attach screenshots. Use code blocks for logs.]

```
(paste logs here)
```

## Additional Notes

[Optional: workarounds, similar issues, etc.]