feat: wire LiteLLM provider into CLI and embeddings

[decko]

Summary

Wire the LiteLLM adapter (from #103) into the CLI and handle embeddings provider dispatch.

Scope (Part 2 of 2)

1. Add litellm to --judge-provider CLI choices
2. Handle embeddings dispatch — keep VertexAIEmbeddings for Vertex, add provider-aware dispatch for LiteLLM models
3. Update --judge-provider help text with examples
4. Add CLI integration tests
5. Update docs (README, getting-started) with per-provider examples

Usage

# Gemini on Vertex AI
raki run --judge --judge-provider litellm --judge-model vertex_ai/gemini-2.5-pro

# OpenAI
raki run --judge --judge-provider litellm --judge-model gpt-4o

# Any LiteLLM-supported model
raki run --judge --judge-provider litellm --judge-model groq/llama-3.1-70b

Files

• src/raki/cli.py — add litellm to provider choices, help text
• src/raki/metrics/ragas/llm_setup.py — embeddings dispatch (if not in feat: add LiteLLM provider adapter for judge metrics #103)
• tests/test_cli.py — integration tests for new provider
• docs/ — usage examples

Depends on

#103 (LiteLLM adapter) must ship first.

[decko]

Implemented in PR #214. Merged to main.