feat(minimax): add MiniMax provider plugin

[octo-patch]

Reason: Add a first-class MiniMax provider plugin so the proxy can serve the MiniMax global and regional endpoints and the MiniMax-M3 and MiniMax-M2.7 models.

What

Adds a new provider plugin under ccproxy/plugins/minimax/, mirroring the
existing provider plugin architecture (config subclass of ProviderConfig,
BaseHTTPAdapter subclass, FastAPI routes, model mappings, and model cards).

• Authentication: static API key sent as a Bearer token; no OAuth
  credential manager or detection service, so the plugin declares no auth
  dependency.
• Regions: a region option (global_en / cn_zh) resolves the OpenAI
  and Anthropic base URLs and the docs root. Explicitly configured URLs are
  preserved.
• Endpoints (mounted under /minimax):
  • POST /v1/chat/completions — OpenAI-compatible.
  • POST /v1/messages — Anthropic-compatible, converted to the OpenAI chat
    protocol through the core format chain.
  • GET /v1/models — configured model metadata.
• Models: default cards and case-insensitive aliases for MiniMax-M3
  and MiniMax-M2.7.
• Registers the minimax entry point, adds a minimax pytest marker, and
  documents a [plugins.minimax] example in config.example.toml.

Tests / checks

• ruff format --check ccproxy/plugins/minimax tests/plugins/minimax
• ruff check ccproxy/plugins/minimax tests/plugins/minimax
• pytest tests/plugins/minimax/unit (11 passed)

[octo-patch]

I updated the MiniMax default model cards with the target context windows, per-million-token pricing, input modalities, and thinking modes, and pushed commit bc663e5. I ran the MiniMax unit tests (11 passed), Ruff checks and format checks, and targeted mypy.