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

[octo-patch]

Summary

Add MiniMax as a dedicated LLM provider, allowing users to configure MiniMax models with provider: "minimax" instead of using the generic openai-compatible provider. The default model is MiniMax-M3 — MiniMax's latest model with 512K context window and up to 128K output.

Changes

• packages/eko-core/src/types/llm.types.ts: Add "minimax" to LLMprovider type union
• packages/eko-core/src/llm/rlm.ts: Add minimax provider branch in getLLM() using createOpenAICompatible with default base URL https://api.minimax.io/v1
• README.md: Add MiniMax example in quickstart code block with MiniMax-M3 as default
• packages/eko-core/test/llm/minimax.test.ts: Add 8 unit tests + 3 integration tests covering M3 / M2.7 / M2.7-highspeed

Usage

const llms: LLMs = {
  default: {
    provider: "minimax",
    model: "MiniMax-M3",  // or MiniMax-M2.7, MiniMax-M2.7-highspeed
    apiKey: "your-minimax-api-key"
  }
};

MiniMax API is OpenAI-compatible, so this integration uses the existing @ai-sdk/openai-compatible package with no new dependencies.

Available Models

Model	Context Window	Best For
MiniMax-M3 (default)	512K tokens, 128K max output	Latest, most capable; supports image input
MiniMax-M2.7	Previous generation	Stable fallback
MiniMax-M2.7-highspeed	Previous generation, lower latency	Fast inference

Test Plan

• 8 unit tests pass (type validation, config, multi-model, multi-provider)
• 3 integration tests included (non-streaming, streaming, tool calling)
• No new dependencies added
• Follows existing provider patterns (modelscope, openrouter)