feat: add LiteLLM support via litellm: model prefix

[RheagalFire]

Summary

Adds LiteLLM support via a litellm: model prefix, letting users route forecasting through 100+ LLM providers (Anthropic, Groq, Bedrock, Vertex AI, DeepSeek, etc.) using the litellm SDK directly - no proxy needed.

Changes

• timecopilot/agent.py - added _resolve_llm() that detects litellm: prefix and creates a pydantic-ai model via litellm.AsyncOpenAI() with drop_params=True
• pyproject.toml - added [litellm] optional dependency (litellm>=1.80.0,<1.87.0)

Example usage

pip install 'timecopilot[litellm]'

# Set your provider's API key
export ANTHROPIC_API_KEY="sk-ant-..."

# Use any litellm model string
timecopilot forecast data.csv --llm litellm:anthropic/claude-sonnet-4-6
timecopilot chat --llm litellm:groq/llama-3.3-70b-versatile

from timecopilot import AsyncTimeCopilot

copilot = AsyncTimeCopilot(llm="litellm:anthropic/claude-sonnet-4-6")
result = await copilot.forecast(df)

Any LiteLLM-supported model string works: anthropic/claude-sonnet-4-6, groq/llama-3.3-70b-versatile, bedrock/anthropic.claude-v2, deepseek/deepseek-chat, etc.

[CLAassistant]

All committers have signed the CLA.

[RheagalFire]

cc @AzulGarza @mcallara

[AzulGarza]

thanks for the contribution @RheagalFire! would it be possible to add a small test in tests/test_live.py?