feat: add MiniMax as cloud LLM provider for inference and processing

[octo-patch]

Summary

Adds MiniMax as an alternative cloud LLM provider alongside the existing Azure OpenAI integration. MiniMax offers powerful models (MiniMax-M2.7, MiniMax-M2.7-highspeed with 204K context) through an OpenAI-compatible API.

Changes

• ragfit/models/minimax_executor.py: MiniMaxExecutor class using OpenAI-compatible SDK with MiniMax's base URL, automatic temperature clamping to (0, 1], and a generate() method for inference compatibility
• ragfit/processing/local_steps/api/minimax.py: MiniMaxChat processing step (mirrors OpenAIChat) for dataset augmentation pipelines
• configs/inference-minimax.yaml: Ready-to-use inference configuration with MiniMax-M2.7
• tests/test_minimax_executor.py: 18 unit tests covering initialization, temperature clamping, chat, generate, and processing step
• tests/test_minimax_integration.py: 3 integration tests (skipped when MINIMAX_API_KEY is not set)
• README.md: Updated with MiniMax provider documentation and usage examples

Usage

Inference:

MINIMAX_API_KEY=your_key python inference.py -cp configs -cn inference-minimax

Processing pipeline:

- _target_: ragfit.processing.local_steps.api.minimax.MiniMaxChat
  inputs: train
  prompt_key: prompt
  answer_key: generated_answer
  instruction: ragfit/processing/prompts/prompt_instructions/qa.txt
  model:
      model: MiniMax-M2.7

Test Plan

• 18 unit tests pass (mocked API)
• 3 integration tests pass with real MiniMax API
• Import and Hydra instantiation verified
• Verify existing tests are not affected