Ollama integration to run models locally

[Winston-Hsiao]

Connect with Ollama to run completions locally (models running on device) no external API requests (to OpenRouter, etc...)

https://ollama.com/

Quality expectations:

1. Test raypaste built locally on feature branch with multiple ollama models running locally (no external network requests being made by raypaste to OpenRouter etc
2. New documentation in README (concise) and dedicated RUN_MODELS_LOCALLY.MD file which is focused on documenting a fully local/on-device raypaste-cli experience.
3. If there are future TODOs that can be labelled as nice-to-have and not required for main feature release, make new Github Issues and document them.