feat: add MiniMax as first-class LLM provider

[octo-patch]

Summary

Add MiniMax as a first-class LLM provider, enabling ii-agent to use MiniMax models (MiniMax-M2.7, MiniMax-M2.7-highspeed) via the OpenAI-compatible API.

Changes

• src/ii_agent/core/config/llm_config.py: Add MINIMAX = "minimax" to APITypes enum
• src/ii_agent/utils/constants.py: Add MiniMax API base URL, default model, and is_minimax_family() helper
• src/ii_agent/llm/minimax.py (new): MiniMaxDirectClient extending OpenAIDirectClient with:
  • Default API base URL (https://api.minimax.io/v1)
  • Temperature clamping to (0.01, 1.0] range (MiniMax rejects 0.0)
  • Automatic stripping of <think>...</think> tags from model responses
• src/ii_agent/llm/__init__.py: Register MiniMax in get_client() factory routing

Tests

• 29 unit tests covering temperature clamping, think-tag stripping, client initialization, factory routing, and constants
• 3 integration tests (require MINIMAX_API_KEY) covering text generation, tool calling, and highspeed model

Usage

from ii_agent.core.config.llm_config import LLMConfig, APITypes

config = LLMConfig(
    model="MiniMax-M2.7",
    api_key=SecretStr("your-minimax-api-key"),
    api_type=APITypes.MINIMAX,
)
client = get_client(config)

Test plan

• All 29 unit tests pass
• All 3 integration tests pass against live MiniMax API
• Verify no regressions in existing provider tests