Add multi-provider LLM support with MiniMax integration

[octo-patch]

Summary

• Add a reusable utils/llm_provider.py factory that lets students switch between OpenAI, MiniMax, Anthropic, Google, and Ollama via LLM_PROVIDER env var
• Auto-detects provider from available API keys (MINIMAX_API_KEY, OPENAI_API_KEY, etc.)
• MiniMax-specific temperature clamping (0 → 0.01) and OpenAI-compatible API routing
• Example scripts: examples/minimax_hello_world.py and examples/multi_provider_example.py
• Updated README to mention MiniMax and document multi-provider usage

Why

The course currently hardcodes ChatOpenAI in all branches. This utility lets students experiment with different LLM providers — including MiniMax (M2.7, M2.5, M2.5-highspeed) — by dropping utils/llm_provider.py into any project branch and changing one env var.

Files changed (9 files, ~523 additions)

File	Description
utils/llm_provider.py	Provider factory with get_chat_model()
examples/minimax_hello_world.py	Standalone MiniMax usage example
examples/multi_provider_example.py	Multi-provider switching demo
tests/test_llm_provider.py	30 unit tests
tests/test_integration_minimax.py	3 integration tests (live API)
README.md	MiniMax in providers list + usage docs
.gitignore	Ignore *.pyc files

Test plan

• 30 unit tests pass (provider detection, temperature clamping, factory, error handling)
• 3 integration tests pass against live MiniMax API
• Verify examples run end-to-end with a MiniMax API key
• Test with other providers (OpenAI, Ollama)

Usage

export LLM_PROVIDER=minimax
export MINIMAX_API_KEY=your-key-here
python examples/minimax_hello_world.py