fix(muyajs): allow CJK ideographs as flanking boundary for emphasis - #4355

Merges 37 upstream commits into the fork, including the pnpm monorepo
conversion (marktext#4302), muya TS rewrite migration (marktext#4314), website package,
and fixes through CJK emphasis boundary (marktext#4355).

Conflict resolutions:
- package.json: adopt upstream monorepo root shell; drop fork's
  root-level deps block; move 4 fork-added deps (diff, @types/diff,
  markdown-it, @types/markdown-it) to packages/desktop/package.json
- build.yml: keep both fork's workflow_dispatch+dynamic-matrix and
  upstream's paths-ignore: packages/muya/**
- wsl.ts, comparisonPane.vue, wsl-paths.spec.ts: accepted git's rename
  detection — files live under packages/desktop/ in the monorepo layout
- docs/dev/FORK_PLAN.md: kept at repo-root docs/dev/ (private fork doc,
  not published website content)
- pnpm-lock.yaml: regenerated via pnpm install

…arktext#4355)

* fix(muyajs): allow CJK ideographs as flanking boundary for emphasis

Closes marktext#4307.

Strict CommonMark 6.2 flanking rules treat a CJK Unified Ideograph as
neither Unicode whitespace nor Unicode punctuation, so `中文**"加粗"**中文`
(and the symmetric Japanese / Korean variants) failed clause (2b) of
canOpenEmphasis / canCloseEmphasis and never opened a strong run. Mirror
the well-known CJK-friendly emphasis extension shipped by Typora, VSCode
markdownlint, etc.: treat CJK Unified Ideographs (BMP + Ext-A + Ext-B
via surrogate pairs + Compatibility Ideographs), Hiragana, Katakana,
Halfwidth Katakana and Hangul Syllables as boundary-equivalent in
clause (2b).

Adds:
- packages/desktop/test/unit/specs/markdown-strong-cjk.spec.ts —
  token-level vitest covering three CJK + inner-punctuation cases plus
  three regression cases that already worked on develop.
- packages/desktop/test/e2e/strong-cjk.spec.ts — Playwright spec
  confirming `<strong>` is emitted in the live WYSIWYG DOM.
- muya/lib/parser shim in src/types/muya.d.ts so the new vitest's
  named-import passes vue-tsc.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* docs(muyajs): mark CJK emphasis widening as a non-standard extension

Strengthen the inline comments in parser/utils.js so it is unambiguous
that the CJK_REG widening of the flanking check is a *deliberate
divergence* from CommonMark spec 6.2, not a spec compliance fix. Note
that GitHub/GFM is strictly conformant by rejecting the same inputs,
and that the extension is additive only (it never rejects emphasis
CommonMark accepts).

No behavioral change.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* fix(muyajs): make flanking-boundary char extraction code-point aware

Address Copilot inline review on PR marktext#4355: CJK_REG contains a
surrogate-pair branch ([\uD840-\uD87F][\uDC00-\uDFFF]) intended to
match non-BMP CJK ideographs (CJK Ext-B onward), but the existing
flanking-boundary readers used charAt / src[i] — single UTF-16 code
unit access — so the regex branch could never fire in practice.

Add lastCodePointChar / codePointCharAt helpers that read a full
Unicode code point (1 or 2 UTF-16 code units) and route every flanking
read in canOpenEmphasis / canCloseEmphasis through them.

Side benefit: PUNCTUATION_REG also contains numerous surrogate-pair
alternations (\uD800[\uDD00-\uDD02\uDF9F\uDFD0] and similar) for
non-BMP Unicode punctuation that were dead for the same reason. Those
spec-defined cases are now correctly tested too.

Adds an extra vitest case using 𠀀 / 𠀁 (CJK Ext-B, U+20000/U+20001)
to lock in the surrogate-pair behavior.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>