feat(models): add [LiteLLM AI Gateway]

[RheagalFire]

Description

• Add LiteLLM AI Gateway as a native model backend, giving outlines users structured generation across 100+ LLM providers (OpenAI, Anthropic, Google, Cohere, Mistral, Bedrock, Azure, Ollama, and more) through a single from_litellm() factory function.
• Adds LiteLLM (sync) and AsyncLiteLLM (async) model classes with streaming support.
• LiteLLMTypeAdapter inherits from OpenAITypeAdapter since LiteLLM uses the same response_format structure.
• Added litellm as optional dependency in pyproject.toml.

Type of change

• New features (non-breaking change).
• Bug fix (non-breaking change).
• Breaking change (fix or feature that would cause existing functionality not to work as expected).