Create extensible framework for file-scoped, targeted / specialized code reviews

[wesm]

I often find myself wanting to analyze files to improve readability, reduce code duplication or redundancy, improve comments, etc. This is something that we should be able to organize with roborev

So basically:

• roborev can ship with built-in "review skills" (for example, refactoring and code duplication / simplification)
• add CLI option to invoke these specialized skills on a single file, multiple files, or a folder considered together
• allow for users to add their own custom prompts / review skills in config.toml so that you aren't stuck with what roborev ships out of the box, or if you need to something particular to your own project, you have an escape hatch to do exactly the kind of targeted analysis. so this would look something like

review_skills = [
  { name = "follow-style", path = "skillz/follow-style.md" }
]

I'll have to think about what this should look like to invoke in the CLI and how it displays in the TUI queue and review views

[wesm]

You can already do this now with your own custom prompts and roborev run but this would provide a cookbook of capabilities rather than having to do your own prompt engineering each time you want to do something that isn't a simple code review.