* Add unit parser benchmark
* Add LRU to memoize function
* Memoize _findUnit
This fixes some performance issues in my heavily-unit-parsing app.
Another idea might be to do an index of reversed unit names, and search
in that, but this is much easier to implement and still should provide
an improvement in the majority of cases (since I'd imagine that most
users tend to prefer a few units at a time, depending on their
application).
* Optimize memoize function
This should be just a little bit faster than before by using Maps, which
have less overhead than javascript objects.
