feat: add MiniMax as LLM provider via OpenAI-compatible API

[octo-patch]

Summary

Add MiniMax AI as a supported LLM provider in R2R's OpenAI completion provider, following the same integration pattern as DeepSeek.

MiniMax offers an OpenAI-compatible API at https://api.minimax.io/v1 with models like MiniMax-M2.7 (latest, 204K context) and MiniMax-M2.5-highspeed (optimized for speed).

Changes

• OpenAI provider (py/core/providers/llm/openai.py): Add MiniMax client initialization (sync + async) with MINIMAX_API_KEY env var, minimax/ prefix routing in both sync and async client selection, and temperature clamping to MiniMax's [0, 1] range
• R2R router (py/core/providers/llm/r2r_llm.py): Add minimax/ to OpenAI-like prefixes for automatic routing
• Config (py/core/configs/minimax.toml): Ready-to-use config for MiniMax-powered R2R deployment
• README (py/README.md): Add MiniMax getting started instructions

Usage

export MINIMAX_API_KEY=your-key
export R2R_CONFIG_NAME=minimax
python -m r2r.serve

Or with the R2R provider (auto-routing):

response = client.retrieval.rag(
    query="What is quantum computing?",
    rag_generation_config={"model": "minimax/MiniMax-M2.7"}
)

Models supported

Model	Context	Best for
minimax/MiniMax-M2.7	204K	Quality, complex reasoning
minimax/MiniMax-M2.5	204K	Balanced performance
minimax/MiniMax-M2.5-highspeed	204K	Speed-optimized

Test plan

• 19 unit tests covering client init, prefix routing, temperature clamping, base args, R2R router dispatch, and config validation
• 3 integration tests (async completion, streaming, JSON mode) with real MiniMax API
• All 22 tests pass