GPT added TODOs could trigger a different flow

[jakethekoenig]

I find when I ask mentat to write very large blocks of code it often adds some TODOs (example attached). I don't think this is necessarily undesirable because:

1. gpt might think better step by step in this way
2. If the method outline it writes is bad it gives the user an opportunity to save tokens/time and ask for higher level changes before implementation

But it would be cool I think if mentat detected TODOs in the code and instead of Apply these changes Y/n/i it said something like: Would you like me to implement the TODOs (Y), Just apply these changes (J) or Throw them out (N).
For example. I'm not sure what the exact phrasing should be.
transcript_20231019_073927.log

[jakethekoenig]

Also side note on our html conversation viewer: github doesn't let you upload html files, presumably because the browser would try to open them and that could be confusing or a security risk. It might be good to also support markdown for easy sharing.