[Feature Request] Configurable max token threshold for auto-compact trigger

Question

**Bug Description**
could you please create a max tokens config so I can use like 450k context? My usecase runs on /loop and I need an auto
compact to run ath this point otherwise I keep running opus 1m with 70% context and the model struggles to logic, ends up becoming lazy

**Environment Info**
- Platform: darwin
- Terminal: ghostty
- Version: 2.1.84
- Feedback ID: 3b6c06f7-1cb7-45aa-82d3-861158b66e90

**Errors**
```json

```

**Note:** Content was truncated.