feat: add MiniMax as first-class LLM provider

[octo-patch]

Summary

Add MiniMax as a native LLM provider, enabling direct API access via MiniMax's OpenAI-compatible endpoint (api.minimax.io/v1) without requiring OpenRouter as an intermediary.

Changes

• intentkit/models/llm.py: Add MINIMAX to LLMProvider enum with is_configured / display_name support; add MiniMaxLLM class (extends LLMModel, uses ChatOpenAI with MiniMax base URL); register in create_llm_model() factory
• intentkit/models/llm.csv: Add MiniMax-M2.7 (intelligence=5, 204K context) and MiniMax-M2.7-highspeed (speed=5, cost-efficient) native model entries
• intentkit/models/llm_picker.py: Update pick_default_model() and pick_summarize_model() to prefer native MiniMax when MINIMAX_API_KEY is configured, with graceful fallback to OpenRouter
• intentkit/config/config.py: Add MINIMAX_API_KEY environment variable
• .env.example: Add MINIMAX_API_KEY= entry

Models

Model	Intelligence	Speed	Context	Input Price	Output Price
MiniMax-M2.7	5/5	3/5	204K	$0.30/M	$1.20/M
MiniMax-M2.7-highspeed	4/5	5/5	204K	$0.07/M	$0.28/M

Backward Compatibility

• Existing OpenRouter-proxied MiniMax models (minimax/minimax-m2.7, minimax/minimax-m2-her) remain unchanged
• Native provider only activates when MINIMAX_API_KEY is set
• No changes to existing providers or their behavior

Test Plan

• 30 unit and integration tests added (all passing)
• test_llm.py: Model loading/filtering with MiniMax key, coexistence with OpenRouter
• test_llm_picker.py: Priority ordering with native MiniMax, fallback behavior
• test_minimax_llm.py: Provider enum, LLM class (base URL, API key, temperature, model override, params), factory routing, model info
• Ruff format and check pass
• All existing tests continue to pass

Co-Authored-By: Octopus liyuan851277048@icloud.com