Add support for more models

[caufieldjh]

This PR will:

• Switch out the current model loading process for one driven by litellm
• Remove the restriction on models to those in models.yaml
• Enable using local models through Ollama (without requiring proxy configuration)

[caufieldjh]

Remaining on this PR:

• Select appropriate embedding model for operation and model provider, or let user specify
• Helper functions for non-openai APIs
  • Also support for Azure endpoints (https://docs.litellm.ai/docs/providers/azure)
  • and for custom URL (https://docs.litellm.ai/docs/providers/openai_compatible)
• Helper functions for running local models
  • Focus on using Ollama (https://docs.litellm.ai/docs/providers/ollama#using-ollama-apichat)
• Docs
• Tests