Optimize quadratic runtime algorithm with tries/suffix trees

[marfox]

if this is a bottleneck you can make it faster via tries or suffix trees. a trie is rather easy to implement but limited to queries like a.startswith(b), while suffix trees are appropriate for full text search (pretty sure there's a library for that if you don't mind additional dependencies)

Originally posted by @e-dorigatti in #405 (comment)