Convert benchmarks to criterion.

[justinlatimer]

Allows benching on stable. Statistics are better, but still quite a lot of variance in the timings from run to run.

[zslayton]

Thanks for the PR! I'm not very familiar with criterion -- I'll try to take a look at this over the weekend.

Statistics are better, but still quite a lot of variance in the timings from run to run.

If it's easy, could you post a before/after of the benchmark output so I can get a better idea of what you mean?

[justinlatimer]

criterion tracks changes across runs - the benchmarks for lifeguard vary a lot across runs (on my machine anyway) that criterion detects the variance as statistically signification.

allocation standard     time:   [1.0051 ms 1.0405 ms 1.0828 ms]
                        change: [+4.4991% +7.4272% +10.486%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 14 outliers among 100 measurements (14.00%)
  5 (5.00%) high mild
  9 (9.00%) high severe