docs(pdftract-5vhp): bring word-boundary-reconstruction.md to v1.0 final-pass
Complete documentation of the adaptive word-boundary algorithm including:
- Initial threshold = 0.25 * font_size
- 20-glyph median adjustment
- 1.5x median formula
- Full Tc/Tw/Tz (character-spacing, word-spacing, horizontal-scaling) corrections
Expanded from 202 lines to 899 lines with:
- Section 3.1: Tc/Tw/Tz formula with explicit parameter table
- Section 3.2: Text-space vs. device-space comparison per plan line 1550
- Section 4: Adaptive algorithm specification (20-glyph window, 1.5× median, outlier exclusion)
- Section 11: Complete pseudo-code (data structures, main loop, detection, threshold computation)
- Section 12: Edge cases (ZWJ, combining marks, CJK, justified text, monospaced, RTL, ligatures, soft hyphens, tabs)
- Section 13: Validation methodology (corpus at tests/fixtures/word-boundary-corpus/, 141 PDFs, 8 categories)
- Section 14: Implementation checklist and references
Closes: pdftract-5vhp
d174725241