Highlight the right characters when folding changes the length

[Jeehut]

Overview

highlightMatchingTokenizedPrefixes(in:locale:) searched a folded copy of the text (diacritics, case and width removed) and applied the found ranges to the original string. Folding changes the length of characters like "é", "ü" or full-width letters, so the highlight slid sideways. It now searches the original string with the folding as comparison options, skips empty tokens, and passes locale on to the tokenization. The @available that only repeated the package minimum is gone.

Old vs. new, side by side:

Text	Search	main highlights	This branch
Café Olé	ole	" Ol"	"Olé"
Ärger über alles	alles	"r all"	"alles"
Ｈｅｌｌｏ World	wor	"ｌ"	"Wor"
Préférences générales	gen	"s g"	"gén"
Meeting mit Jürgen über Budget	budget	"r Budg"	"Budget"

Identical results for ASCII, case differences, "ß", half-width kana, CJK, emoji, an empty search, double spaces and Turkish with locale: tr. No case was found in which the new code is worse.

Out of scope: both versions highlight only the first occurrence per token, even in the middle of a word ("lock" in "unlock"). That is a separate change.

Split out of #7.

How this was verified

StringExtTests covers plain matches and the folding cases above as a parameterized test.

Red, with main's StringExt.swift:

✘ "Ärger über alles" / "uber"  expected ["über"]
✘ "Ｈｅｌｌｏ World" / "wor"     expected ["Wor"]
✘ "Café Olé" / "ole"           expected ["Olé"]
✘ "Ärger über alles" / "alles" expected ["alles"]
✘ Test run with 2 tests in 1 suite failed with 4 issues.

Green, this branch:

✔ Test run with 2 tests in 1 suite passed.