test(qa): take #899's rows into the Linux LibreOffice fidelity baseline

[DemchaAV]

Why

#899 moves five documents: cv-orange_ops, cv-serif_headline, invoice-workspace, proposal-editorial and proposal-northline. The Linux LibreOffice baseline still holds those five as they stood before #899. #898 says its baseline matches the docx-fidelity artifact of #899's last CI run (38571103664) byte for byte. The merged files differ from that artifact in exactly these five documents' rows and match it everywhere else.

Every CI run since then measures the corpus the way #899's run did; the artifacts of #906 and #907 match it byte for byte. Because the five documents now stand nearer the page than the baseline says, the DOCX fidelity job passes. It would also pass a change that undid #899 on Linux.

What changed

libreoffice-linux.tsv and libreoffice-linux-lines.tsv now match the docx-fidelity artifact of #907's CI run (38794754386) byte for byte, summary and lines. That run measured 2.5-dev at 43adc8336.

Only the rows of the five documents change.

• Lines that moved by more than 0.05pt: 45 nearer the page, 2 further. Both further lines belong to cv-orange_ops and moved by 0.05pt each.

• Summary rows:

  Document	Before	After
  proposal-editorial	median 0.21pt, p90 0.65pt	median 0.20pt, p90 0.35pt
  proposal-northline	median 0.17pt, p90 0.48pt	median 0.13pt, p90 0.42pt
  cv-serif_headline	p90 0.65pt	p90 0.59pt
  cv-orange_ops	p90 0.36pt	p90 0.35pt

  invoice-workspace's summary row stays the same; two of its lines move nearer.

Verification

• ./mvnw -B -ntp clean verify → BUILD SUCCESS (render-docx 1628, qa 1820).
• The committed files are the artifact's bytes: git show HEAD:<file> | cmp - <artifact file> reports no difference for either file.
• This PR's own DOCX Fidelity job measures the same tree. If the baseline is right, it reports no line further from the page.

Lane: test (qa fidelity baseline), no code change.