fix(cli): count a contraction as one word in line_unread - #6346

What

readingWords() keeps an apostrophe between letters inside a word run, so can't, can’t, don't and it's count as one word each.

Why

WORD_RUNS stopped at ' and ’, so every contraction counted as two words and line_unread set a longer hold floor than the line's real word count. "a tooth can't get in" counted 6 words instead of 5.

Related work

Refs #6338. This covers the contractions. Hyphenated words still count per part: Intl.Segmenter splits rock-and-roll at the hyphens even inside one run, and for a hold time three parts read about as long as three words, so I left that as is. The docs pointer about the captions track from the same issue isn't in here either.

How

The run regex now takes ['’] followed by more letters or numbers: /[\p{L}\p{M}\p{N}%$]+(?:['’][\p{L}\p{M}\p{N}]+)*/gu. A quote that doesn't sit between letters ('quoted', a lone ') still isn't part of a run. The segmenter keeps can't as one word-like segment, so the CJK path is unchanged.

Test plan

• Unit tests added/updated: new case in craftAudit.test.ts (straight and typographic apostrophe, plus a quoted word). It fails on main (expected 6 to be 5) and passes here; the file's 21 tests pass.
• Manual testing performed
• Documentation updated (if applicable)
• Comments follow CONTRIBUTING.md "Comments": they say why, not what, and a bug fix says what the code must do and how to reproduce the bug

oxlint and oxfmt --check are clean on both files.