feat(subtitles): keep text subtitle styling and placement to the host

[superuser404notfound]

Summary

Re #233. Text subtitle styling (bold, italic, underline, strikeout, colour, font face and size) and the placement a cue asks for now reach the host instead of being stripped at the parser.

What changed

• One parser, not one per format. libavcodec converts every text subtitle format into an ASS event line before the engine sees it, so the markup was already on AVSubtitleRect.ass: SRT through ff_htmlmarkup_to_ass, WebVTT through its own tag table, teletext because the decoder already ran txt_format=ass. cleanASSBody stripped every {...} block and the colour parser handled \c and ignored the rest. coloredRuns becomes styledRuns, a full override parser, and SRT, WebVTT, teletext and ASS light up together.
• Model. SubtitleTextRun gains isBold, isItalic, isUnderlined, isStruckThrough, fontName, fontSize. SubtitleCue gains placement (SubtitleTextPlacement: ASS numpad alignment plus an optional anchor normalized to [0, 1] like SubtitleImage.position). New initializer parameters are defaulted, so no call site changes.
• \pos normalization uses a declared PlayResX/Y when the ASS header has one, otherwise libavcodec's 384x288 default, which is what its synthesised lines use.
• Unconditional, no gate. An unstyled cue still arrives as .text with the same string, so a host handling only that case sees no change; .richText was already reachable through teletext.
• PiP compositor no longer flattens rich text into one white font. Per-run attributed line (colour, face, ASS-relative size, real CoreText bold/italic traits, underline, drawn rule for strikeout since CoreText has no attribute) and placement honoured.
• Lookalike tags are left alone: \be/\bord are not \b, \iclip is not \i, \shad is not \s, \fscx/\fsp are not \fs.
• The AU live tv #107 teletext blank-line fold moved from the run parser into teletextBody, where it belongs: only libzvbi's row joining produces those, and a blank line in an ASS or SRT cue can be deliberate.

Not covered

WebVTT cue settings (line, position, align, size, vertical) never arrive: libavcodec does not convert them, webvttdec.c says so in its file header. VTT inline styling works, VTT positioning needs an upstream patch. SRT positioning needs AV_PKT_DATA_SUBTITLE_POSITION side data most files lack.

Test plan

• Issue233ASSOverrideStyleTests (21): attributes, toggles, multi-tag blocks, ASS weights, font name/size and their resets, lookalike tags, \r, colour regressions, cue-level \an/\pos, play-res normalization and header parsing, SRT- and VTT-shaped lines, body selection, escapes, and the teletext fold boundary.
• Issue233CompositorStylingTests (12): placement maths for all nine alignments with and without an explicit position, font trait and size resolution, render smoke tests.
• TeletextColourRunsTests moved onto the new API with every AU live tv #107 assertion preserved.
• 1210 tests green, strict-concurrency=complete clean, tvOS and iOS Simulator BUILD SUCCEEDED.

Checklist

• CHANGELOG.md updated
• README.md updated
• Commit messages follow Conventional Commits
• The fix lives in the engine, not in a host-side workaround
• Public API changes are intentional and documented