LiteLLM Integration + Type Fixing + Usage Tracking + Testing

[sidjha1]

• Remove OpenAIModel and just have a LM class that uses LiteLLM
• Add mypy and pre-commit and get mypy to pass (also add mypy to CI).
• Add usage tracking with lm.print_total_usage()
• Add ollama tests with llama3.2:3b
• Verify that code runs with vLLM
• Merge in RM and Reranker refactoring from Refactor RM and Reranker #28

Notes:

• I had to do a bit of random/weird plumbing to get mypy to pass since our types were pretty messed up