Better support for LM Studio local provider/models #508

(Formalizing the sentiment in this comment: #195 (comment)) as a new feature support request.

I hate to say this, but the "other" (previous, pre-forked) opencode had better support and easier setup for LM Studio... although it was clunky and didn't work so well (so I understand the new, more explicit JSON configuration approach backed by model specs.)

That said, I'd really love it if the initial "blank-slate" auth/login/token screen offered LM Studio (maybe prompting for the local URL with default in lieu of login or token), then just used the /models endpoint to lookup what supported models are available to populate the /models popup, as with OpenRouter or other providers. It's kinda a pain to set it up and keep the config files updated as I make changes to my model selection in LM Studio, which seems to happen a lot these days as models are released very often.

At the very least, additional LM Studio - related documentation in the quickstart documentation would be extremely helpful. It took me a long time to find the helpful comment referenced at the top of this request, and I was stuck because I didn't realize specifying the model separately was actually required (not just the provider... kinda redundant?).

Other than that, I really like this fork... it's coming along nicely!