This module implements a trie data structure. The term "trie" comes from the
word retrieval, but is generally pronounced like "try". A trie is a tree
structure (or directed acyclic graph), the nodes of which represent letters
in a word. For example, the final lookup for the word 'bob' would look
something like $ref->{'b'}{'o'}{'b'}{'00'} (the 00 being an end marker).
Only nodes which would represent words in the trie exist, making the structure
slightly smaller than a hash of the same data set.

The advantages of the trie over other data storage methods is that lookup times
are O(1) WRT the size of the index. For sparse data sets, it is probably not as
efficient as performing a binary search on a sorted list, and for small files,
it has a lot of overhead. The main advantage (at least from my perspective) is
that it provides a relatively cheap method for finding a list of words in a
large, dense data set which begin with a certain string.