#### Don't split lines between Japanese kana & COMBINING KATAKANA-HIRAGANA (SEMI-)VOICED SOUND MARK (#15411 by @tats-u)

This PR fixes [#15410](https://github.com/prettier/prettier/issues/15410).

Japanese (semi-)voiced kana characters can be split into two code points. For example, the following hiragana character /ka/ can be represented as:

が (U+304C) → か (U+304B) + ゙ (U+3099) → が (U+304C U+3099)

Most users do not use or meet expressions like this except for file paths in macOS. However, there are some characters that can only be represented in this way. Some Japanese text that have to tell /ŋa̠/ (there are not a few Japanese that do not use it these days though) from the common /ga/ use a expression ["か゚" (U+304B U+309A)](https://en.wiktionary.org/wiki/%E3%81%8B%E3%82%9A).

```markdown
nasalか゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚
```

The above Markdown is formatted as in Prettier 3.0:

```markdown
nasalか゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚か゚き゚く゚け
゚こ゚
```

The semi-voiced sound mark goes to the next line but it is not correct. By this PR, the source Markdown is now formatted as:

```markdown
nasalか゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚か゚き゚く゚け゚こ゚か゚き゚く゚
け゚こ゚
```

The semi-voiced sound mark now keeps attached to the hiragana "け".