the concept of an extra-width per code-unit was obsolete at least since apps
supported unicode with its different normalization forms, diacritical marks,
surrogate-pairs, non-printing characters such as ZWJ/ZWNJ/RLM, etc.  so of
course modern engines like CoreText don't aid this typographical crime.

The fix here extends the CTLayout::GetTextBreak() method to handle the obsolete
semantic of per code-unit extra-widths by successively approximating the number
of involved code-units.