Introduce benchmarks

[sdboyer]

benchmarks are gonna be particularly crucial. in general, two things are worth measuring:

1. relatively low N benchmarks, useful for longitudinal comparisons as changes come in and performance comparisons across different analogous implementations
2. varying range of N benchmarks, useful for giving a loose picture on the performance characteristics of each implementation's individual methods as graph size increases