TUI runs out of context with gpt-5.5 despite auto-compact config

Question

### What happened
In Codex CLI TUI, a long chat repeatedly failed with:

```text
Codex ran out of room in the model's context window. Start a new thread or clear earlier history before retrying.
```

### Config
```toml
model = "gpt-5.5"
# model_context_window = 400000
# model_auto_compact_token_limit = 300000
```

### Expected
The TUI should auto-compact before reaching the model context limit, or explain why auto-compaction did not run or did not apply.

### Notes
This happened repeatedly in th