feat: enhance layout preprocessing and unify image scaling proposal

Backend changes:
- Add image scaling configuration for PP-Structure processing
- Enhance layout preprocessing service with scaling support
- Update OCR service with improved memory management
- Add PP-Structure enhanced processing improvements

Frontend changes:
- Update preprocessing settings UI
- Fix processing page layout and state management
- Update API types for new parameters

Proposals:
- Archive add-layout-preprocessing proposal (completed)
- Add unify-image-scaling proposal for consistent coordinate handling

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

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

backend/app/services/pp_structure_enhanced.py

@@ -7,10 +7,14 @@ This module provides enhanced PP-StructureV3 processing that extracts all
 
 import logging
 from pathlib import Path
-from typing import Dict, List, Optional, Tuple, Any
+from typing import Dict, List, Optional, Tuple, Any, TYPE_CHECKING
 import json
 import gc
 
+# Import ScalingInfo for type checking (avoid circular imports at runtime)
+if TYPE_CHECKING:
+    from app.services.layout_preprocessing_service import ScalingInfo
+
 # Optional torch import for additional GPU memory management
 try:
     import torch
@@ -81,7 +85,8 @@ class PPStructureEnhanced:
         image_path: Path,
         output_dir: Optional[Path] = None,
         current_page: int = 0,
-        preprocessed_image: Optional[Image.Image] = None
+        preprocessed_image: Optional[Image.Image] = None,
+        scaling_info: Optional['ScalingInfo'] = None
     ) -> Dict[str, Any]:
         """
         Analyze document with full PP-StructureV3 capabilities.
@@ -93,10 +98,13 @@ class PPStructureEnhanced:
             preprocessed_image: Optional preprocessed PIL Image for layout detection.
                                If provided, this is used for PP-Structure prediction,
                                but original image_path is still used for cropping images.
+            scaling_info: Optional ScalingInfo from preprocessing. If image was scaled
+                         for layout detection, all bbox coordinates will be scaled back
+                         to original image coordinates for proper cropping.
 
         Returns:
             Dictionary with complete structure information including:
-            - elements: List of all detected elements with types and bbox
+            - elements: List of all detected elements with types and bbox (in original coords)
             - reading_order: Reading order indices
             - images: Extracted images with metadata
             - tables: Extracted tables with structure
@@ -184,7 +192,7 @@ class PPStructureEnhanced:
                 # Process parsing_res_list if found
                 if parsing_res_list:
                     elements = self._process_parsing_res_list(
-                        parsing_res_list, current_page, output_dir, image_path
+                        parsing_res_list, current_page, output_dir, image_path, scaling_info
                     )
                     all_elements.extend(elements)
 
@@ -247,13 +255,15 @@ class PPStructureEnhanced:
         parsing_res_list: List[Dict],
         current_page: int,
         output_dir: Optional[Path],
-        source_image_path: Optional[Path] = None
+        source_image_path: Optional[Path] = None,
+        scaling_info: Optional['ScalingInfo'] = None
     ) -> List[Dict[str, Any]]:
         """
         Process parsing_res_list to extract all elements.
 
         Args:
             parsing_res_list: List of parsed elements from PP-StructureV3
+            scaling_info: Scaling information for bbox coordinate restoration
             current_page: Current page number
             output_dir: Optional output directory
             source_image_path: Path to source image for cropping image regions
@@ -285,11 +295,28 @@ class PPStructureEnhanced:
 
             # Ensure bbox has 4 values
             if len(layout_bbox) >= 4:
-                bbox = layout_bbox[:4]  # [x1, y1, x2, y2]
+                bbox = list(layout_bbox[:4])  # [x1, y1, x2, y2]
             else:
                 bbox = [0, 0, 0, 0]  # Default if bbox missing
                 logger.warning(f"Element {idx} has invalid bbox: {layout_bbox}")
 
+            # Scale bbox back to original image coordinates if image was scaled
+            # This is critical for proper cropping from original high-resolution image
+            if scaling_info and scaling_info.was_scaled and bbox != [0, 0, 0, 0]:
+                scale_factor = scaling_info.scale_factor
+                bbox = [
+                    bbox[0] * scale_factor,  # x1
+                    bbox[1] * scale_factor,  # y1
+                    bbox[2] * scale_factor,  # x2
+                    bbox[3] * scale_factor   # y2
+                ]
+                if idx == 0:  # Log only for first element to avoid spam
+                    logger.info(
+                        f"Scaled bbox to original coords: "
+                        f"{[round(x, 1) for x in layout_bbox[:4]]} -> {[round(x, 1) for x in bbox]} "
+                        f"(factor={scale_factor:.3f})"
+                    )
+
             # Extract content (check multiple possible keys)
             content = (
                 item.get('content', '') or