fix: add proper coordinate scaling from OCR space to PDF space

Problem:
- OCR processes images at smaller resolutions but coordinates were being used directly on larger PDF canvases
- This caused all text/tables/images to be drawn at wrong scale in bottom-left corner

Solution:
- Track OCR image dimensions in JSON output (ocr_dimensions)
- Calculate proper scale factors: scale_w = pdf_width/ocr_width, scale_h = pdf_height/ocr_height
- Apply scaling to all coordinates before drawing on PDF canvas
- Support per-page scaling for multi-page PDFs

Changes:
1. ocr_service.py:
   - Add OCR image dimensions capture using PIL
   - Include ocr_dimensions in JSON output for both single images and PDFs

2. pdf_generator_service.py:
   - Calculate scale factors from OCR dimensions vs target PDF dimensions
   - Update all drawing methods (text, table, image) to accept and apply scale factors
   - Apply scaling to bbox coordinates before coordinate transformation

3. test_pdf_scaling.py:
   - Add test script to verify scaling works correctly
   - Test with OCR at 500x700 scaled to PDF at 1000x1400 (2x scaling)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

backend/app/services/ocr_service.py

@@ -332,6 +332,7 @@ class OCRService:
                 total_valid_regions = 0
                 all_layout_data = []
                 all_images_metadata = []
+                all_ocr_dimensions = []
 
                 for page_num, page_image_path in enumerate(image_paths, 1):
                     logger.info(f"Processing PDF page {page_num}/{len(image_paths)}")
@@ -363,6 +364,14 @@ class OCRService:
                         if page_result.get('images_metadata'):
                             all_images_metadata.extend(page_result['images_metadata'])
 
+                        # Store OCR dimensions for each page
+                        if page_result.get('ocr_dimensions'):
+                            all_ocr_dimensions.append({
+                                'page': page_num,
+                                'width': page_result['ocr_dimensions']['width'],
+                                'height': page_result['ocr_dimensions']['height']
+                            })
+
                 # Calculate overall average confidence
                 avg_confidence = total_confidence_sum / total_valid_regions if total_valid_regions > 0 else 0.0
 
@@ -407,11 +416,18 @@ class OCRService:
                     'processing_time': processing_time,
                     'timestamp': datetime.utcnow().isoformat(),
                     'total_pages': len(image_paths),
+                    'ocr_dimensions': all_ocr_dimensions if all_ocr_dimensions else None,
                 }
 
             # Get OCR engine (for non-PDF images)
             ocr_engine = self.get_ocr_engine(lang)
 
+            # Get the actual image dimensions that OCR will use
+            from PIL import Image
+            with Image.open(image_path) as img:
+                ocr_width, ocr_height = img.size
+                logger.info(f"OCR processing image dimensions: {ocr_width}x{ocr_height}")
+
             # Perform OCR
             logger.info(f"Processing image: {image_path.name}")
             # Note: In PaddleOCR 3.x, use_angle_cls is set during initialization, not in ocr() call
@@ -480,6 +496,10 @@ class OCRService:
                 'markdown_content': markdown_content,
                 'processing_time': processing_time,
                 'timestamp': datetime.utcnow().isoformat(),
+                'ocr_dimensions': {
+                    'width': ocr_width,
+                    'height': ocr_height
+                }
             }
 
             logger.info(