feat: add Novita AI as LLM provider

[Alex-wuhu]

Summary

• Adds Novita AI as an OpenAI-compatible LLM provider via LiteLLM
• Registers qwen3-embedding-0.6b token limit (8192) in EMBEDDING_MODEL_LIMITS
• Adds NOVITA_API_KEY detection to health_check.py env validation
• Adds Novita AI configuration example to .env.example
• Adds Novita AI setup section to docs/installation_and_configuration.rst

Configuration

Users can configure Novita AI by adding to their .env:

# Chat model
NOVITA_API_KEY=<your-novita-api-key>
OPENAI_API_BASE=https://api.novita.ai/openai
CHAT_MODEL=openai/moonshotai/kimi-k2.5

# Embedding model
LITELLM_PROXY_API_KEY=<your-novita-api-key>
LITELLM_PROXY_API_BASE=https://api.novita.ai/openai
EMBEDDING_MODEL=litellm_proxy/qwen/qwen3-embedding-0.6b

Novita AI uses the OpenAI-compatible endpoint, so it routes through LiteLLM's existing openai provider path with no new dependencies.

Test plan

• Set NOVITA_API_KEY and run rdagent health_check — env_check should pick up the key and test chat/embedding connectivity
• Verify openai/moonshotai/kimi-k2.5 model resolves correctly via LiteLLM with OPENAI_API_BASE=https://api.novita.ai/openai
• Verify litellm_proxy/qwen/qwen3-embedding-0.6b embedding call succeeds with Novita base URL

---

📚 Documentation preview 📚: https://RDAgent--1369.org.readthedocs.build/en/1369/