This object simplifies the interactions with tags. The user only ever deals
with the raw tag values. The parser will unescape the tags and store them in
the tags objects. Likewise, the serializer will escape the tags as it outputs
them.

Right now the tags object works like an ordered hash table. This is mostly for
unit tests, but is probably fine going forward. That said, I'm not particularly
fond of the implementation, but it does work so it's good enough for now.

Ran the parser and tags tests under valgrind.

Called in the turtles as well.