[Feature Request] Per-model spend caps and token allowance allocation

**Bug Description**
I would love it if I can set a spend cap per model, for example, I would allow Fable to use 50% of the current session token allowence, that'd be great for me and many other devs.

**Environment Info**
- Platform: darwin
- Terminal: androidstudio
- Version: 2.1.268
- Feedback ID: 2ef8781e-9c72-49a2-b715-d858a741e12d

**Errors**
```json
[]
```

The issue names no files, tests, or entry points. Begin by locating the existing session token allowance and model-selection behavior, then confirm the intended configuration and acceptance criteria for per-model caps and allocation.

由索引模型根据 Issue 内容生成。