fix(convert): preserve PDF→PPTX line layout and font names

[nelsonduarte]

Problem

User report after v1.13.8: "já reconhece o texto mas fica tudo desformatado".

The previous fix made PPTX output editable but used one textbox per block, with each line of the block as a paragraph inside. PowerPoint's textframe re-flowed those paragraphs based on font metrics, ignoring the per-line y-coordinates from the PDF. Multi-line blocks collapsed visually — headings + body that were spaced apart in the PDF stacked together in PowerPoint.

Fix

Switch from per-block to per-line textboxes:

• Each line from page.get_text("dict") gets its own add_textbox at the line's bbox, so PowerPoint has nothing to re-flow vertically and every line stays at its original PDF y position.
• Spans within a line stay as runs in a single paragraph — inline formatting (bold/italic mid-sentence) is preserved.
• word_wrap = False per textbox since each holds exactly one line; this stops PowerPoint from breaking a line across two visual lines if its glyph width metrics differ from the PDF's by a hair.

Also propagate the PDF span's font name to the run, stripping the 6-letter subset prefix that fitz reports (e.g. ABCDEF+Arial → Arial). PowerPoint substitutes a similar face if the exact font isn't installed, but at least the canonical name lets the substitution be reasonable instead of always landing on the theme default Calibri.

Test plan

• Smoke test on Ubuntu 26.04 + Py3.14.4: a 1-page PDF with text at y=100/300/700, sizes 22/14/10, produces 3 separate textboxes with vertical order preserved, sizes preserved exactly, font name = Helvetica:

  'Title at top'    y=76.35  size=22.0  name=Helvetica
  'Middle line'     y=284.95 size=14.0  name=Helvetica
  'Bottom footnote' y=689.25 size=10.0  name=Helvetica
  

• Live test in PowerPoint with the user's original problematic PDF: lines should land at their original positions; selection / editing still works on each line independently.

🤖 Generated with Claude Code