feat (services): groq, mlx-lm, huggingface, mistral-inference, anthro…

…pic, gemini can't promise i can keep maintaining this, as i'm working on a startup also this does not test the CALL functions of the models, only their creation - [ ] test llm calls for the new services - [ ] find a better word than "garbage" for services with "closed output" - [ ] set up ci.yml to run github actions to automatically test all the things (especially mlx)