Queued GitHub Copilot commands merged and use wrong model.

1. Issue the command `/compact`.
2. While the conversation is being compacted, issue a custom prompt `/foo` with Model A selected, and press ``. The prompt will be queued.
3. While the conversation is still being compacted, issue a custom prompt `/bar` with Model B selected, and press ``. The prompt will also be queued.

**Expected:** When the conversation is finished being compacted, prompt `/foo` will be submitted using Model A; and when it is finished prompt `/bar` will be submitted using Model B.

**Actual:** When the conversation is finished being compacted, prompt `/foo` and prompt `/bar` are combined into a single prompt and submitted using Model A.

This is blatantly wrong on many fronts:

* The prompts should have been submitted separately because they were queued separately; if I would have wanted them combined, I would have submitted them combined.
* The result of combining the prompts means that **`/bar` wasn't even recognized as a custom prompt** and instead was apparently submitted as the literal test `/bar` in the body of the combined prompt. (At least `/bar` was not displayed as being recognized as a custom prompt.)
* The second prompt, `/bar`, needed a more powerful model, Model B. If I would have wanted to use Model A for `/bar`, I would have selected Model A instead of Model B.

```
Version: 1.130.0 (user setup)
Commit: 1b6a188127eeaf9194f945eb6eb89a657e93c54c
Date: 2026-07-22T14:55:04Z
Electron: 42.6.0
ElectronBuildId: 14623276
Chromium: 148.0.7778.280
Node.js: 24.18.0
V8: 14.8.178.38-electron.0
OS: Windows_NT x64 10.0.26200
```