feat: add MiniMax as alternative LLM provider

[octo-patch]

Summary

Add MiniMax M2.7 and M2.7-highspeed as alternative chat model providers via their OpenAI-compatible API.

Changes

• config.py: Add MiniMax-M2.7 (1M context) and MiniMax-M2.7-highspeed to SUPPORTED_GPT_MODELS; add open_ai_base_url config field; temperature clamping for MiniMax (must be > 0); is_minimax_model() helper
• ai.py: Pass base_url to OpenAI() client constructor; tiktoken fallback to cl100k_base for non-OpenAI models; fix streaming ChoiceDelta attribute access
• config.example.json: Add open_ai_base_url field
• readme.md / readme.zh.md: Add MiniMax configuration documentation (EN + ZH)
• tests/test_minimax.py: 20 unit tests + 3 integration tests

Usage

{
  "open_ai_key": "your-minimax-api-key",
  "open_ai_base_url": "https://api.minimax.io/v1",
  "open_ai_chat_model": "MiniMax-M2.7"
}

Embeddings still use OpenAI embedding API. MiniMax is chat-only.

Test plan

• 20 unit tests pass (model definitions, config parsing, base_url, temp clamping, tiktoken fallback)
• 3 integration tests pass (mock + live MiniMax API calls, streaming)
• Existing OpenAI models still work