feat: add MiniMax as cloud LLM provider

[octo-patch]

Summary

Add MiniMax M2.7 and M2.7-highspeed as cloud LLM providers alongside existing local models (ChatGLM, BELLE, Vicuna, InternLM, Yuan2). This enables users without GPU hardware to use the knowledge-based QA system via MiniMax's cloud API.

Changes

• minimax_llm.py (new): ChatMiniMaxLLM class extending LangChain LLM, using MiniMax OpenAI-compatible API (https://api.minimax.io/v1) with temperature clamping to (0, 1] and think-tag stripping
• config.py: Added MiniMax model entries (MiniMax-M2.7, MiniMax-M2.7-highspeed) to llm_model_dict; fixed missing comma syntax error
• app.py: Added MiniMax detection in init_model_config() — uses ChatMiniMaxLLM for cloud inference, skipping local model loading
• requirements.txt: Added openai>=1.0.0 dependency
• README.md: Added MiniMax models to supported models table with usage instructions

Usage

export MINIMAX_API_KEY="your-api-key"
python app.py
# Select MiniMax-M2.7 or MiniMax-M2.7-highspeed from the model dropdown

Test Plan

• 21 unit tests covering initialization, API calls, temperature clamping, history management, think-tag stripping, stop tokens
• 5 integration tests verifying config syntax, app dispatch logic, README content
• 3 API integration tests with real MiniMax API (basic completion, highspeed model, Chinese Q&A)
• All 29 tests pass (28/29 with network; 1 transient timeout)