Same design as a BytesTrie, but mapping any UnicodeString (any sequence of 16-bit units) to 32-bit integer values. This can use somewhat simpler code because there are more bits to work with in each unit, and it is probably more appropriate and faster than a BytesTrie for collation contractions/prefixes, CJK dictionaries, and maybe for use with Unicode strings in general when it is not known that we work with a small script or mostly with ASCII.
The code and data structure are quite similar to the BytesTrie. In general, larger units are used to store larger values and deltas in single units than possible in a BytesTrie, and fewer variable-length units are needed in all cases.
In addition, some of the bits of match-nodes (linear-match and branch nodes) are used for intermediate values (small values or most significant bits), rather than separate intermediate-value nodes in a BytesTrie. Larger intermediate values have one or two units following the match node head, then followed by the match node’s contents.