OCTO-11116

[OlteanuRares]

Summary

• Parse WebVTT cue settings (position, line, size, align, vertical) into structured Layout objects — integer line numbers and percentage values both convert to viewport percentages; vertical:rl|lr stored as WritingDirectionEnum
• Parse STYLE blocks: extract ::cue CSS rules (font-style, font-weight, text-decoration, color, background-color) and resolve them onto <c.classname> spans at read time with correct cascade (::cue < ::cue(.class))
• Add WritingDirectionEnum to geometry.py and extend Layout with writing_direction field
• Filter ::cue-prefixed style keys from DFXP and SAMI writers to prevent invalid selectors in output

Production safety

These changes do not affect current production flows (SCC → all output formats):

• SCC reader produces no STYLE blocks or cue settings — the new parsing code is never triggered
• CaptionSet.get_style("::cue") returns {} for non-VTT sources, so style resolution is a no-op
• The serialized() tuple adds a trailing None for writing_direction on SCC/DFXP sources — equality/hash behavior is unchanged
• DFXP and SAMI writers explicitly skip ::cue-prefixed style keys

[github-actions]

🟡 PR Compliance Review

Risk Level: MEDIUM

• Compliance Issues: 1 (0 critical)
• Regressions: 0

REVIEW REQUIRED - Address issues before merging

Full report available in workflow artifacts