feat: add MiniMax provider support

[octo-patch]

Summary

• Add MiniMax chat model provider via the OpenAI-compatible endpoint (https://api.minimax.io/v1)
• Add MINIMAX_API_KEY environment variable support and a make_lm() helper on EvolutionConfig that auto-routes MiniMax model strings to the correct base URL and API key
• Add --use-minimax CLI flag to evolve_skill.py for easy one-flag switching
• Update SyntheticDatasetBuilder and LLMJudge to use config.make_lm() so all LM calls benefit from MiniMax routing
• Document MiniMax usage in README with model table and usage examples
• Add 16 unit tests covering constants, env var loading, model routing, temperature enforcement, and pass-through behavior for non-MiniMax models

Supported models

Model ID	Description
MiniMax-M2.7	Peak performance — default
MiniMax-M2.7-highspeed	Same performance, lower latency

Usage

export MINIMAX_API_KEY=your_key_here

# Shorthand
python -m evolution.skills.evolve_skill --skill github-code-review --use-minimax

# Explicit
python -m evolution.skills.evolve_skill \
    --skill github-code-review \
    --optimizer-model minimax/MiniMax-M2.7 \
    --eval-model minimax/MiniMax-M2.7-highspeed

API Reference

• Chat (OpenAI Compatible): https://platform.minimax.io/docs/api-reference/text-openai-api

Test plan

• pytest tests/core/test_minimax_config.py — 16 new tests, all passing
• pytest tests/ (excluding pre-existing broken test_constraints.py) — 139 passed, 0 failed

[jlicerio]

LGTM! Approved by Hermes Agent.