feat(ocr): Add image dimension limit and progress callback for PDF OCR

[ErixWong]

This PR enhances the PDF OCR converter with two new features:

1. Image Dimension Limiting (default 1500px)

• Prevents oversized images from being sent to LLM Vision APIs
• Configurable via MARKITDOWN_MAX_IMAGE_DIMENSION environment variable
• Uses LANCZOS resampling for quality preservation
• Backward compatible - defaults to 1500px if not specified

2. Progress Callback Support

• Allows applications to track PDF processing progress
• Reports progress percentage and current operation
• Works for both regular PDF parsing and full-page OCR fallback
• Optional parameter - no callback means no progress reporting

Benefits

• Better performance with large PDFs containing high-res images
• Improved user experience for long-running operations
• Reduced API costs by resizing images before sending
• Backward compatible - all new parameters are optional

Testing

• Tested with various PDF files including scanned documents
• Verified progress callback works correctly
• Confirmed image resizing maintains OCR quality

Fixes: N/A (feature enhancement)