Microbenchmarks

[Benson Margulies]

Post 0.5, what do people think about populating some microbenchmarks?
If we're going to consider dumping collections for fastutil, it might
be good to have a measurement framework. I've had some success with
japex.