feat: Add Groq provider support

[Shanzita]

Summary

• Add Groq as a new LLM provider
• Support 5 models: llama-3.3-70b-versatile, llama-3.1-8b-instant, llama-3.1-70b-versatile, mixtral-8x7b-32768, gemma2-9b-it
• Groq offers very fast inference, useful for latency-sensitive routing

Changes

• llm_inference/model_inference.py: Add _call_groq() method and model mappings
• model_cost/model_cost.json: Add Groq model pricing
• universal_model_names.py: Add Groq model names
• .env.example: Add GROQ_API_KEY

Test plan

• CI passes (pre-commit, mypy)
• Manual test with Groq API key (optional)

Notes

Groq uses OpenAI-compatible API, implementation follows existing patterns (similar to DeepSeek/Perplexity).

[yl231]

Apologies for the long silence here, @Shanzita — picking up maintainer triage. This adds Groq provider support (additive, ~77 lines) and looks reasonable in principle, but it's ~5 months behind main and will need a rebase before we can run CI on it. Is this still something you'd like to land? If so, please rebase onto the latest main and I'll review + run the checks. If the models you were targeting are now reachable via an existing provider, we can close it instead.

[Shanzita]

Hi @yl231, thanks for picking this up! Yes, I'd still like to land this. I've rebased the branch onto the latest main and resolved the conflicts — the change is still purely additive (~77 lines). Pre-commit and the MyPy 3.9–3.12 checks pass locally on the rebased branch, so CI should be green. Ready for review whenever you get a chance.