fix(ocr): improve text layer alignment with width-based font sizing

- Create new hocr-transform.ts utility for parsing hOCR output
- Add line-aware text processing with baseline and rotation support
- Implement width-based font size calculation to match word bounding boxes
- Fix text selection not covering full characters issue
- Add proper type definitions for OcrLine, OcrPage, WordTransform
- Support RTL languages and CJK word break handling

src/js/types/ocr-pdf-type.ts

@@ -1,10 +1,46 @@
 export interface OcrWord {
-    text: string;
-    bbox: { x0: number; y0: number; x1: number; y1: number };
-    confidence: number;
+  text: string;
+  bbox: { x0: number; y0: number; x1: number; y1: number };
+  confidence: number;
 }
 
 export interface OcrState {
-    file: File | null;
-    searchablePdfBytes: Uint8Array | null;
+  file: File | null;
+  searchablePdfBytes: Uint8Array | null;
+}
+
+export interface BBox {
+  x0: number; // left
+  y0: number; // top (in hOCR coordinate system, origin at top-left)
+  x1: number; // right
+  y1: number; // bottom
+}
+
+export interface Baseline {
+  slope: number;
+  intercept: number;
+}
+
+export interface OcrLine {
+  bbox: BBox;
+  baseline: Baseline;
+  textangle: number;
+  words: OcrWord[];
+  direction: 'ltr' | 'rtl';
+  injectWordBreaks: boolean;
+}
+
+export interface OcrPage {
+  width: number;
+  height: number;
+  dpi: number;
+  lines: OcrLine[];
+}
+
+export interface WordTransform {
+  x: number;
+  y: number;
+  fontSize: number;
+  horizontalScale: number;
+  rotation: number;
 }