Character offsets count UTF-32 code units.

Implementation note: these are the same as Unicode codepoints, so this PositionEncodingKind may also be used for an encoding-agnostic representation of character offsets.