Add Transliterate SE5 plugin

[niksedk]

New plugin: converts subtitle text between scripts, letter by letter (no translation, offline).

Languages

• Cyrillic ↔ Latin: Serbian, Montenegrin (ś/ź), Macedonian (Scientific or ASCII system), Uzbek (oʻ/gʻ, any apostrophe accepted)
• To Latin: Russian (Readable or ISO 9), Ukrainian (official 2010), Belarusian (BGN/PCGN style), Bulgarian (official 2009), Kazakh (2021 alphabet), Greek (ELOT 743), Korean (Revised Romanization with sound-change rules)

Kept as-is: tags (<i>, {\an8}), URLs/e-mails, letter case (Љубав → Ljubav, ЉУБАВ → LJUBAV), text already in the target script. When going back to Cyrillic it also leaves words with q/w/x/y, Roman numerals and an editable Keep in Latin list. Handles Serbian split digraphs (injekcija, nadživeti).

UI (Avalonia + Material Design icons, follows the SE theme)

• Language list with colored code tiles; the detected language is marked (guessed from the letters used, e.g. ђ/ћ/џ = Serbian, є/ї = Ukrainian, Serbian Latin with đ/ć)
• Direction card with script tiles (Ж / Ω / 한 / A) and a swap button, next to a live example
• System buttons for Russian/Macedonian, remembered per language
• All/Selected lines, checkable before/after cards, OK = one undo step

Uses Subtitle.Paragraphs (keeps ASSA fields), falls back to SubRip for older SE.

Tests: 61 xunit tests (every language, detection, keep rules, view model). Checked the window headless in light and dark, Serbian both directions, Korean and the empty state.

Not verified by a native speaker: the Kazakh 2021 alphabet details (і = ı, И = İ casing) and the Belarusian system choice are the least certain; Korean sound changes cover the common rules only.

After merge: release via the workflow, then add the plugin to se5-plugins.json and the README.

🤖 Generated with Claude Code