feat: OCR umlaut confusion correction + bold detection via stroke-width

- Add umlaut confusion rules (i→ü, a→ä, o→ö, u→ü) to _spell_fix_token
  for German text — fixes "iberqueren" → "überqueren" etc.
- Add _detect_bold() using OpenCV stroke-width analysis on cell crops
- Integrate bold detection in both narrow (cell-crop) and broad (word-lookup) paths
- Add is_bold field to GridCell TypeScript interface
- Render bold text in StepGroundTruth reconstruction view

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

admin-lehrer/components/ocr-pipeline/StepGroundTruth.tsx

@@ -389,6 +389,7 @@ export function StepGroundTruth({ sessionId, onNext }: StepGroundTruthProps) {
                         height: `${cell.bbox_pct.h}%`,
                         color: '#1a1a1a',
                         fontSize: `${fontSize}px`,
+                        fontWeight: cell.is_bold ? 'bold' : 'normal',
                         fontFamily: "'Liberation Sans', 'DejaVu Sans', Arial, sans-serif",
                         display: 'flex',
                         alignItems: 'center',