feat: add MiniMax as OpenAI-compatible LLM provider

[octo-patch]

Summary

• Add MiniMax as a first-class OpenAI-compatible LLM provider
• MiniMax offers high-performance models (M2.5, M2.7) with up to 1M token context windows via an OpenAI-compatible API at https://api.minimax.io/v1
• Follows the existing provider pattern (FireworksAI, Perplexity, TogetherAI)

Changes

• MiniMaxApiConfiguration.ts: New API configuration class extending BaseUrlApiConfigurationWithDefaults, uses MINIMAX_API_KEY env var
• OpenAICompatibleFacade.ts: Added MiniMaxApi() factory function alongside existing providers
• index.ts: Export MiniMaxApiConfiguration
• MiniMaxApiConfiguration.test.ts: 8 unit tests covering API config, streaming, and text generation (using MSW mocks)
• MiniMaxApiConfiguration.integration.test.ts: 3 integration tests (skipped without MINIMAX_API_KEY)
• README.md: Added MiniMax to provider lists and usage example

Usage

import { generateText, openaicompatible } from "modelfusion";

const text = await generateText({
  model: openaicompatible
    .ChatTextGenerator({
      api: openaicompatible.MiniMaxApi(),
      model: "MiniMax-M2.5",
    })
    .withTextPrompt(),
  prompt: "Write a short story about a robot learning to love:",
});

Available models: MiniMax-M2.5, MiniMax-M2.5-highspeed (204K context), MiniMax-M2.7 (1M context)

Test plan

• All 8 unit tests pass (API config validation, streaming text, non-streaming text generation)
• All 161 existing tests continue to pass (no regressions)
• Integration tests pass with valid MINIMAX_API_KEY (auto-skipped in CI)