When importing captions with missing/invalid bbox data, elements are now
arranged in a 4-column grid (row/column layout) instead of a diagonal
pattern that overlapped after 6 elements.
- Change fallback positioning from modulo-6 diagonal to row/column grid
- Update both Python backend and JavaScript editor
- Add comprehensive test for 12+ elements with no overlap
- Fixes visual collision when pasting captions with many unbounded regions
Co-authored-by: Cursor Agent <cursoragent@cursor.com>