feat: add MiniMax as a new LLM engine

[octo-patch]

Summary

• Add MiniMax as a new LLM engine option (-e minimax)
• Uses Python stdlib urllib.request to maintain the zero-dependency design
• Supports MiniMax-M2.5 (default, 204K context) and MiniMax-M2.5-highspeed models
• Requires MINIMAX_API_KEY environment variable
• Update README with MiniMax usage examples and documentation

Usage

export MINIMAX_API_KEY=your-api-key

# Use default MiniMax-M2.5 model
python3 autoprompt.py seed.txt criteria.md -e minimax --target 9.0

# Use high-speed variant
python3 autoprompt.py seed.txt criteria.md -e minimax -m MiniMax-M2.5-highspeed

Why MiniMax?

MiniMax provides powerful cloud LLM models with a generous 204K context window, which is especially useful for evolving longer text artifacts. Unlike the existing CLI-based engines, MiniMax requires no CLI tool installation — just an API key and Python stdlib.

Test plan

• Verify Python syntax compiles without errors
• Verify CLI help shows minimax as an engine option
• Verify error handling when MINIMAX_API_KEY is not set
• Manual test with a valid MiniMax API key