[Epic] Advanced Text, Auto-fit & Internationalization

[MHoroszowski]

Problem

The text-run API is incomplete in three orthogonal directions. Run-level features: superscript/subscript (scanny/python-pptx#1045), strikethrough (#574, #339), highlight, baseline shift, character spacing (spc), and kerning (kern) are absent or partial. Layout features: text columns (numCol), vertical text (vert), per-run column-spacing — <a:bodyPr> round-trips them but no Python API exposes them. Auto-fit & overflow: TextFrame.fit_text exists but crashes on long words (#168, #416) and there is no overflow-detection API to query whether content fits before saving (#715, #936). Internationalization: RTL paragraphs (#751) are not supported, and the <a:latin>/<a:ea>/<a:cs> typeface trio means setting Font.name only fixes Latin script — Asian fonts and complex-script fonts cannot be set independently (#343, #768, #420).

Sub-features

• Font.superscript / Font.subscript (boolean → emits baseline= percentage)
• Font.strike enum (none / single / double)
• Font.highlight color setter
• Font.character_spacing (spc=) and Font.kerning (kern=)
• Independent Font.latin / Font.east_asian / Font.complex_script setters (existing Font.name continues setting only <a:latin>)
• TextFrame.columns = N and TextFrame.column_spacing (<a:bodyPr numCol= colSpace=>)
• TextFrame.text_direction enum (horizontal / vert / vert270 / wordArtVert / eaVert / mongolianVert)
• Paragraph.rtl = True for right-to-left text (Arabic, Hebrew, Persian)
• TextFrame.will_overflow() / TextFrame.overflow_info() returning a structured report (closes scanny/python-pptx#1114)
• TextFrame.shrink_text_to_fit() — eager font-size autofit without relying on PowerPoint to recompute on open (closes scanny/python-pptx#1107)
• Fix fit_text long-word crash (closes #168)

Prior art

• Open PRs: scanny/python-pptx#1046 sub/superscript (lxrichert, clean). scanny/python-pptx#601 older sub/super. scanny/python-pptx#606 strikethrough (gtessa, recently active). scanny/python-pptx#112 older strike. scanny/python-pptx#1107 shrink_text_to_fit (BastienGimbert). scanny/python-pptx#1114 overflow detection (saggl, 1382 LOC tested). scanny/python-pptx#751 RTL paragraphs (comatan96). scanny/python-pptx#416 fit_text long-word crash fix.
• Forks:
  • Force1ess/python-pptx — strikethrough at run + textframe-level font attrs.
• User issues this would close: #100, #114, #133, #134, #144, #168, #285, #339, #343, #378, #420, #485, #514, #525, #537, #563, #574, #675, #684, #710, #715, #720, #765, #768, #773, #830, #836, #884, #936, #969, #970, #1026, #1045.
• Already shipped in this fork (excluded): bullet/numbered lists on paragraphs (#100, #114, #655, #807).
• OOXML: DrawingML §21.1.2.3.7 <a:rPr> (u, strike=sngStrike/dblStrike, baseline, spc, kern), §21.1.2.1.1 <a:bodyPr> (numCol, colSpace, vert, rtlCol, anchor), §21.1.2.2.7 <a:pPr> (rtl), §21.1.2.3.5 <a:latin>/<a:ea>/<a:cs> typeface trio.
• Code paths: src/pptx/text/text.py, src/pptx/text/fonts.py, src/pptx/oxml/text.py, src/pptx/text/layout.py (TextFitter).

Acceptance criteria

• 35+ unit tests + 10 behave scenarios across the sub-features.
• TextFrame.will_overflow() correctly reports overflow status for content >= 110% of frame height across 3 sample fonts.
• A presentation authored with Paragraph.rtl = True and Arabic text renders right-to-left in PowerPoint.
• East-Asian font set via Font.east_asian = "MS Gothic" round-trips correctly.
• Backward compat: existing Font.name continues setting only the Latin slot.

Effort: L

Largest text-API surface in the roadmap. Phase A: run-level (sub/super, strike, highlight). Phase B: typography + i18n (latin/ea/cs trio, RTL). Phase C: layout (columns, vertical) and overflow detection.