feat: add MiniMax provider support

[octo-patch]

Summary

This PR adds MiniMax LLM provider support to LangKit's extensible LLMInvocationParams architecture.

Changes

• langkit/openai/openai.py: Added two new provider classes:
  • MiniMaxDefault — uses MiniMax-M2.7 model (default, peak performance)
  • MiniMaxHighSpeed — uses MiniMax-M2.7-highspeed model (faster, more agile)
  • Added create_minimax_chat_completion() helper supporting both legacy (v0) and modern (v1+) openai library versions
• langkit/openai/__init__.py: Exported MiniMaxDefault and MiniMaxHighSpeed
• langkit/tests/test_minimax.py: Added 11 unit tests covering model names, temperature defaults, copy behavior, completion calls, and custom base URL support

Usage

from langkit.openai import MiniMaxDefault, MiniMaxHighSpeed
from langkit.openai.openai import Conversation

# Set MINIMAX_API_KEY in environment
provider = MiniMaxDefault()
conv = Conversation(provider)
result = conv.send_prompt("Hello!")

Configuration

Env Var	Description	Default
MINIMAX_API_KEY	MiniMax API key (required)	—
MINIMAX_BASE_URL	Custom base URL	https://api.minimax.io/v1

Key Constraints

• Temperature default is 1.0 (MiniMax API requires range (0.0, 1.0])
• Uses OpenAI-compatible API endpoint

API Reference

• Chat (OpenAI Compatible): https://platform.minimax.io/docs/api-reference/text-openai-api