Add MiniMax as LLM provider with thinking model middleware

[octo-patch]

Summary

• Add MiniMax M2.7 as a supported LLM provider using @ai-sdk/openai OpenAI-compatible adapter
• Include AI SDK wrapLanguageModel middleware that handles two MiniMax M2.7 quirks:
  • Downgrades json_schema response format to json_object (MiniMax doesn't support json_schema), embedding the schema in the system prompt
  • Strips <think>...</think> tags from both generate and stream responses (M2.7 is a thinking model)
• Add MiniMax to README provider list with setup instructions
• Add examples/minimax.ts with complete working example

Test plan

• 6 unit tests for provider configuration and LLMScraper initialization (tests/minimax.test.ts)
• 3 integration tests for scraping, streaming, and markdown format (tests/minimax-integration.test.ts) — requires MINIMAX_API_KEY env var, auto-skipped otherwise
• All 9 tests passing

Files changed

File	Description
README.md	Add MiniMax to features list and provider setup section
examples/minimax.ts	Full example with middleware for thinking model support
tests/minimax.test.ts	6 unit tests for provider config and scraper init
tests/minimax-integration.test.ts	3 integration tests (scraping, streaming, markdown)