feat(ai-model-router-v2): add MiniMax provider to cloud registry

[octo-patch]

Reason: The ai-model-router-v2 cloud registry did not include the MiniMax provider, so its text models and regional API endpoints were unavailable.

Changes

• Add the MiniMax text models to the ai-model-router-v2 cloud registry (skill/modules/detector.py):
  • MiniMax M3 — 1,000,000-token context; text, image and video input; adaptive/disabled thinking; pricing (USD per million tokens) input 0.6, output 2.4, cache read 0.12.
  • MiniMax M2.7 — 204,800-token context; text input; always-on thinking; pricing input 0.3, output 1.2, cache read 0.06, cache write 0.375.
• Add per-region API endpoints exposing both an OpenAI-compatible and an Anthropic-compatible base URL, surfaced via ModelDetector.get_cloud_endpoints():
  • Global (global_en): https://api.minimax.io/v1 and https://api.minimax.io/anthropic.
  • China (cn_zh): https://api.minimaxi.com/v1 and https://api.minimaxi.com/anthropic.
• Extend ModelInfo with optional metadata fields (context_window, input_modalities, thinking, pricing_usd_per_million_tokens, openai_base_url, anthropic_base_url) to carry these parameters; existing entries are unchanged.
• Document the provider, model parameters and regional endpoints in SKILL.md, and note MiniMax support in the catalog README.md.
• Add tests/test_minimax_registry.py validating the registry entries and endpoints.

Checks

• python3 -m py_compile skill/modules/detector.py skill/core/router.py tests/test_minimax_registry.py — passed.
• python3 -m unittest test_minimax_registry -v — 5 tests passed.