> Sure, these automata are “wide”, with lots of branches, but they’re also shallow, since they run on UTF-8 encoded characters whose maximum length is four and average length is much less

I would consider splitting this task into two:

- extracting the next Unicode code unit

- determining whether it’s in the code class

For the second, instead of using an automaton, one could use a perfect hash (https://en.wikipedia.org/wiki/Perfect_hash_function). That could make that part branch-free.

Is that a good idea?