feat: add MiniMax as 4th LLM provider (M2.7, 204K context)

[octo-patch]

Summary

Add MiniMax as a first-class LLM provider alongside OpenAI, Anthropic, and Qwen. MiniMax offers the M2.7 model with 204K context window and an OpenAI-compatible API, making integration seamless via ChatOpenAI with a custom base URL — no new dependencies required.

Changes

• default_config.py: Add minimax_api_key config field
• trading_graph.py: Add MiniMax to _get_api_key(), _create_llm() (with temperature clamping to 0.01–1.0), and update_api_key()
• web_interface.py: Add MiniMax to provider validation, API key management, key status, and error handling endpoints
• templates/demo_new.html: Add MiniMax to LLM provider dropdown and API key input group in the web UI
• README.md / README_CN.md: Add MiniMax environment variable setup and acknowledgement
• tests/: 19 unit tests + 3 integration tests covering config, factory, temperature clamping, API key lifecycle, provider switching, and web endpoints

MiniMax Models

Model	Context	Use Case
MiniMax-M2.7	204K tokens	Agent + Graph LLM (default)
MiniMax-M2.7-highspeed	204K tokens	Faster inference

Usage

export MINIMAX_API_KEY="your_minimax_api_key_here"
python web_interface.py
# Select "MiniMax" from the LLM Provider dropdown

Or via config:

config = {
    "agent_llm_provider": "minimax",
    "graph_llm_provider": "minimax",
    "agent_llm_model": "MiniMax-M2.7",
    "graph_llm_model": "MiniMax-M2.7",
    "minimax_api_key": "your-key"
}
tg = TradingGraph(config=config)

Test Plan

• 19 unit tests pass (config, factory, temp clamping, API key, web endpoints, provider switching)
• 3 integration tests pass (real MiniMax API calls with M2.7-highspeed)
• Manual: select MiniMax in web UI dropdown, enter API key, run analysis

Notes

• MiniMax uses the existing ChatOpenAI class with openai_api_base="https://api.minimax.io/v1" — zero new dependencies
• Temperature is clamped to (0.01, 1.0] per MiniMax API requirements
• All existing OpenAI/Anthropic/Qwen functionality remains unchanged