I have made an unicode tr39 confusable detection and skeleton implementation, ex_confusables
This library is able to compare two strings if they are visually confusable as described in Unicode® Technical Standard #39: Both strings get transformed into a skeleton format before comparing them. The skeleton is generated by normalizing the string (NFD), replacing confusable characters, and normalizing the string again.
Also did some profiling between char list and binary string implementation. The interesting part is char list is faster than binary. The result is here.
Feel free to open any issues you encounter in the Github repo and try it!