Add MiniMax as built-in LLM provider

[octo-patch]

Summary

Adds MiniMax as a built-in LLM provider. MiniMax offers OpenAI-compatible API endpoints, so this integration leverages the existing openai SDK with a custom base_url — no extra dependencies needed.

Models registered

Model	Description
MiniMax-M2.7	Latest flagship model (204K context)
MiniMax-M2.7-highspeed	Fast variant of M2.7
MiniMax-M2.5	Previous-gen thinking model
MiniMax-M2.5-highspeed	Fast variant of M2.5

Usage

import ell

# Auto-registered when MINIMAX_API_KEY env var is set
@ell.simple(model="MiniMax-M2.7")
def hello(name: str):
    """You are a helpful assistant."""
    return f"Say hello to {name}."

# Or with explicit client
import openai
client = openai.Client(base_url="https://api.minimax.io/v1", api_key="...")

@ell.simple(model="MiniMax-M2.5", client=client)
def chat(prompt: str):
    """You are a helpful assistant."""
    return prompt

Changes (5 files, +478 lines)

• src/ell/models/minimax.py — Model registration with auto-detection via MINIMAX_API_KEY
• src/ell/models/__init__.py — Import minimax module
• tests/test_minimax_provider.py — 16 unit tests + 3 integration tests
• examples/providers/minimax_ex.py — Usage example
• docs/src/core_concepts/models_and_api_clients.rst — Add MiniMax to provider list

Test plan

• 16 unit tests for model registration and OpenAI provider compatibility
• 3 integration tests with real MiniMax API (chat completion, streaming, custom client)
• All 87 existing tests pass without regressions

Design notes

• Follows the same pattern as xai.py — OpenAI client with custom base_url
• No new provider class needed since MiniMax is fully OpenAI-compatible
• Silently skips auto-registration when MINIMAX_API_KEY is not set