Add optional local Ollama support with strict provider isolation

[TurboRx]

#98 (comment)

you told me to add this so i added it. i didn't do property testing on it because i don't have powerful hardware to test local/self-hosted models.

[AgustinSRG]

Thank you for the contribution. I tested it with my local Ollama installation and it works properly.

Even if you have less powerful hardware, some models may run on it. I recommend this website to check what may run on your GPU: https://www.canirun.ai/

Also, Ollama will run the model with RAM + CPU (~ 40 times slower or more) if your GPU does not support it, so you can even run an LLM on a CPU with the drawback of being very slow.

Before merging, please fix the linter issues reported by the CI. You can check in local by running npm test. You can also fix the trivial ones automatically by running npm run fix-test

[TurboRx]

Fixed lint.