XLSX: account for thick row borders in printed tracks

[developer0hye]

Summary

• read both row-boundary flags from raw worksheet XML because public umya-spreadsheet v2 drops thickTop; upstream support is proposed in fix: preserve row thickTop MathNya/umya-spreadsheet#359 and merged into the workspace's v2 fork in fix: preserve row thickTop umya-spreadsheet#8
• add Excel's fixed 1pt reservation for an automatic row's current thickTop and the preceding automatic row's thickBot, after the font-dependent row-track mapping
• keep customHeight rows unchanged because their declared tracks already include those boundaries, with a regression against the native monthly-budget chart size
• add 150 DPI fix evidence showing all nine SH107 table baselines and eight baseline pitches move onto the native Excel grid

Related issue

Related: #1228

Testing

• red regression before the fix: SH107's nine printed row heights were all Some(17.0) instead of 18/19/17/19/19/17/17/17/17
• red custom-height regression before narrowing the rule: two 40pt rows grew to 41pt and 42pt
• cargo fmt --check
• cargo test -p office2pdf --lib -- parser::xlsx::tests — 275 passed
• cargo test -p office2pdf --lib -- parser::xlsx::row_boundaries::tests — 2 passed
• cargo test -p office2pdf --test xlsx_fixtures — 206 passed
• cargo clippy --workspace --all-targets -- -D warnings
• cargo publish --dry-run -p office2pdf --allow-dirty — packaged library verified against public crates.io dependencies
• python3 scripts/compare_layout.py ... --page 1 --noise-floor 0.5 --audit — after: 0.00pt mean/worst baseline delta, 0.00pt pitch delta, no large shifts
• python3 scripts/compare_render.py ... --page 1 --dpi 150 --artifacts-dir ... --lines
• magick compare -metric AE -fuzz 5% ... — page-1 AE fell from 132,379 pixels (6.29253%) before to 35,937 (1.70824%) after
• final-build 150 DPI page-1 raster is byte-identical to the committed after.jpg source raster
• git diff --check
• visual evidence validator — no errors

Visual impact

• No rendered PDF change
• Rendered PDF change or visual evidence added
• Reason: Automatic rows with thick top/bottom boundary flags now reserve the native fixed points in their printed tracks.

Visual audit

• Issue: XLSX: a row declaring thickTop prints at the bare recomputed track, where Excel grows it 2pt #1228
• Fixture: tests/fixtures/xlsx/SH107-9-x-9-Formatted-Table.xlsx
• Page(s): page 1 for fix evidence; output page 2 also inspected and confirmed inkless
• Renderer and DPI: native Excel-for-Mac PDF export and pdftoppm, 150 DPI
• Evidence mode: fix
• New follow-up issues found in this audit: None
• Model vision findings: The full page, 5% diff, and matched header/table/thick-row crops show the before table compressing upward after each thick boundary. After the fix all row rules, baselines, fills, dark double outlines, and orange underlines align with GT; the inkless extra page, black Col9 label, and sub-point right-aligned text offset remain separately tracked below.
• GT: assets/bugfixes/issue-1228/gt.jpg
• Before: assets/bugfixes/issue-1228/before.jpg
• After: assets/bugfixes/issue-1228/after.jpg

Visual comparison

GT	Before	After

Required inspection

• Rendered all evidence at 150 DPI or higher
• Stored progressive JPEG quality 86 assets with metadata stripped
• Used Codex/Claude vision to inspect the full GT/output pages, diff, and matched crops
• Inspected matched region crops at full resolution
• Ran compare_layout.py --audit and dispositioned every large text-instance shift
• Ran the 5% fuzz pixel-difference sweep
• Inventoried hairlines and border dash styles
• Inventoried font weight, italic, and underline emphasis

The hairline inventory covers every light-blue cell rule, the dark double outlines around Col5/1.5, and the orange rules under Col8/1.8; all are present at the matching positions and none uses a dash pattern. The emphasis inventory confirms the bold header labels, italic Col6/1.6, bold 1.4, and underlined orange runs. The remaining page-1 diff clusters are text colour/raster placement covered by #1230 and #1233, plus accepted antialiasing; #1229 covers the inkless second page outside the evidence page.

Deviation audit

Check	Result
Page count/order	Remaining: #1229 tracks the inkless second page; the native export has one page.
Element presence	Matches GT on page 1 — every label, value, fill, rule, and outline is present.
Position/size	Remaining: #1233 tracks sub-point horizontal placement of right-aligned runs; all nine baselines and eight row pitches are fixed and match GT exactly.
Rotation/flip	No deviation observed — the fixture contains no rotated or flipped elements.
Fill	Matches GT — all header, band, and highlighted-cell fills align.
Stroke/border	Matches GT — all light-blue hairlines, dark double outlines, and orange underlines are present at matching positions and no dash styles occur.
Shape outline geometry	No deviation observed — the page has no shapes beyond the audited cell borders.
Text content	Matches GT — all 81 cell values and nine header labels are present.
Font family/weight/style	Matches GT — header bold, Col6/1.6 italic, 1.4 bold, and orange underlined emphasis are preserved.
Text color	Remaining: #1230 tracks the black Col9 header label that is white in GT.
Alignment	Remaining: #1233 tracks the right-aligned trailing-advance residual; vertical alignment is fixed.
Line/paragraph spacing	Fixed — native and output baseline pitches are 19/17/19/19/17/17/17/17pt.
Clipping/overflow	Remaining: #1229 tracks value-less trailing columns that create the blank overflow page; no page-1 clipping was observed.

Checklist

• Commits include a Signed-off-by line
• PR scope contains one root cause
• Remaining visual deviations each reference an open issue