Add MiniMax as a first-class LLM and embedding provider

[octo-patch]

Summary

• MiniMaxLLM: Chat completions via OpenAI-compatible API with temperature clamping (0.01–1.0) and automatic <think> tag stripping for reasoning models
• MiniMaxEncoder: Native embedding API for embo-01 model (1536 dimensions) with sync/async support via requests/aiohttp
• Registered MiniMax in EncoderType enum, AutoEncoder factory, EncoderDefault, and LLM __init__ exports

Changes

File	Description
semantic_router/llms/minimax.py	MiniMaxLLM class extending BaseLLM
semantic_router/encoders/minimax.py	MiniMaxEncoder class extending DenseEncoder
semantic_router/llms/__init__.py	Register MiniMaxLLM
semantic_router/encoders/__init__.py	Register MiniMaxEncoder + AutoEncoder factory
semantic_router/schema.py	Add MINIMAX to EncoderType enum
semantic_router/utils/defaults.py	Add MINIMAX defaults (embo-01, MiniMax-M2.5)
tests/unit/llms/test_llm_minimax.py	17 unit tests for LLM
tests/unit/encoders/test_minimax.py	16 unit tests for encoder
tests/integration/test_minimax_integration.py	6 integration tests
docs/integrations/minimax.md	Integration documentation
docs/integrations/integrations.md	Add MiniMax to provider list
README.md	Add MiniMax to integrations list

Test plan

• 33 unit tests passing (17 LLM + 16 encoder)
• Integration tests for LLM chat completions (sync + async)
• Integration tests for encoder initialization
• Embedding integration tests require sufficient API rate limits

MiniMax Models

Type	Model	Notes
Chat	MiniMax-M2.5	Default, 204K context
Chat	MiniMax-M2.5-highspeed	Fast inference
Embedding	embo-01	1536 dimensions