feat: add MiniMax as first-class LLM provider (M3 default)

[octo-patch]

Summary

Add MiniMax as a first-class LLM provider in AdalFlow's model client system, with MiniMax-M3 as the default model.

• New MiniMaxClient extending OpenAIClient — MiniMax's API is fully OpenAI-compatible
• Base URL: https://api.minimax.io/v1
• API key: read from MINIMAX_API_KEY environment variable
• Supported models: MiniMax-M3 (default; 512K context, up to 128K output, image input), MiniMax-M2.7, MiniMax-M2.7-highspeed
• Registered in model_client/__init__.py via LazyImport (uses openai SDK dependency)

Changes

File	Change
adalflow/components/model_client/minimax_client.py	New MiniMaxClient class with docs and example usage (default MiniMax-M3)
adalflow/components/model_client/__init__.py	Register MiniMaxClient via LazyImport + add to __all__
tests/test_minimax_client.py	12 unit tests covering init, sync/async clients, Generator integration, serialization (assertions target MiniMax-M3)

Usage

from adalflow.core import Generator
from adalflow.components.model_client import MiniMaxClient

generator = Generator(
    model_client=MiniMaxClient(),
    model_kwargs={
        "model": "MiniMax-M3",
        "temperature": 0.7,
        "stream": False,
    }
)

output = generator(prompt_kwargs={"input_str": "Hello!"})

Test plan

• All 12 new unit tests pass (pytest tests/test_minimax_client.py -v)
• Existing client tests unaffected (test_deepseek_client.py, test_sambanovaclient.py, test_openai_client.py all pass)
• Follows the exact same pattern as DeepSeekClient, XAIClient, SambaNovaClient