Feat/add litellm provider

[RheagalFire]

Adds LiteLLM as a provider with 48 models across 8 providers.

LiteLLM is an AI gateway/proxy that routes to 100+ LLM providers through a single OpenAI-compatible endpoint. Users who run a LiteLLM proxy server can connect it to opencode and access any model behind it.

Provider config

• npm: @ai-sdk/openai-compatible
• env: LITELLM_API_KEY
• api: http://localhost:4000/v1 (default proxy URL)

Models added (48)

• Anthropic (9): Claude Opus 4.6/4.5/4.1, Sonnet 4.6/4.5, Haiku 4.5, 3.7 Sonnet, 3.5 Sonnet v2, 3 Haiku
• OpenAI (20): GPT-5.5 through GPT-4 Turbo, o3/o4 series, Codex variants
• Google (5): Gemini 2.5/2.0/1.5 Flash & Pro
• DeepSeek (2): Chat, Reasoner
• Mistral (5): Large, Small, NeMo, Codestral, Devstral
• Cohere (2): Command A, Command R+
• xAI (3): Grok 3, 3 Mini, 2
• Groq (2): Llama 3.3 70B, 3.1 8B

All models use extends to inherit definitions from existing providers. bun validate passes.