Replicate auggie's prompt enhancer feature

### Describe the feature or problem you'd like to solve

Writing good prompt and predicting how the model is going to interpret them is hard

### Proposed solution

The prompt enhancer from Augment Code is fantastic.

It not only helps generate better prompts, but because the model is actually evaluating the prompt, you also get an understanding of how the model interpreted your initial prompt.

This gives you the ability to clarify your prompt and correct any mistaken assumptions made by the mode's evaluation of the prompt

Pro Tip: I had "Question?" to the input to the prompt enhancer and it often asks questions that will help it understand your intent better

### Example prompts or workflows

In the auggie CLI, prompt enhancer is involved by `^p`. This pairs really well with the request in issue #78

### Additional context

My original prompt
```
For our basline http over ssh performance, I think it would be better to use openssh's local forwaring instead of a raw http session. Let's make that change

questons?
```

Prompt enhancer ->
```
For our baseline HTTP-over-SSH performance testing, I want to modify the test implementation to use OpenSSH's built-in local port forwarding (`ssh -L`) instead of our current approach that creates raw HTTP sessions through SSH tunnels programmatically.

Specifically:
1. Replace the current SSH tunnel implementation that uses librssconnect's `SshSession::createForwardL()`
2. Instead, spawn an external OpenSSH process with local port forwarding (e.g., `ssh -L local_port:target_host:target_port`)
3. Then run the HTTP performance tests against the locally forwarded port
4. This will provide a more realistic baseline that represents how users typically use SSH tunneling in practice

The goal is to establish whether our librssconnect SSH tunneling performance is competitive with standard OpenSSH local forwarding, or if there are optimization opportunities.

Questions:
1. Do you want to keep the current tunneling tests as well for comparison, or replace them entirely?
2. Should we use the same SSH keys and configuration that our current tests use with the OpenSshTestServer?
```