feat(extracted-text): enforce OCR segment validation with bbox/type/reading_order

Title (semantic)

feat(extracted-text): enforce OCR segment validation with bbox/type/reading_order

Description

This merge request implements OCR-specific extracted text segment validation while keeping ASR behavior compatible.

What was implemented

• Extended OCR segment support in extracted text schema with:
  • bbox
  • type
  • reading_order
• Added OCR-only validation rules:
  • bbox, type, reading_order, start, end are required per segment
  • end == start + 1 (page-index rule)
  • bbox must have 4 values and valid geometry (x2 > x1, y2 > y1)
  • type must be one of:
    • Text, Title, Caption, Table, Picture, Formula, Section-header, List-item, Page-header, Page-footer, Footnote
  • segment reading_order must be in proper order
• Preserved ASR compatibility:
  • OCR-only fields are not mandatory for extraction_type="asr"
  • existing ASR temporal consistency validation remains

Files changed

• app/schemas/extracted_text.py
• tests/unit/schemas/test_extracted_text_ocr_segments.py
• tests/integration/api/test_extracted_text_ocr_validation.py

Checklist

• The feature has been fully implemented.
• Tests for the new feature are included and passing.
• User documentation/guides have been updated (if applicable).
• Impact on existing functionality has been considered.

Related Issue(s)

Closes #107 (closed)