feat(ocr): Add environment variable support for OCR configuration

[ErixWong]

This PR adds the ability to configure OCR services via environment variables, following the 12-factor app principles.

Changes

• Add from_env() class method to LLMVisionOCRService for creating instances from environment variables (MARKITDOWN_OCR_API_KEY, MARKITDOWN_OCR_API_BASE, MARKITDOWN_OCR_MODEL)
• Update plugin registration to support three-tier configuration priority:
  1. Explicit kwargs (llm_client, llm_model, llm_prompt)
  2. Environment variables
  3. No OCR service (text-only extraction)

Benefits

• No need to hardcode API keys in code
• Easier deployment in containerized environments
• Backward compatible - existing code continues to work

Testing

• Tested with local environment variables
• Verified backward compatibility with existing code

Fixes: N/A (feature enhancement)