Test-drive Criterion.rs

[garious]

See: https://japaric.github.io/criterion.rs/book/criterion_rs.html

We gain:

• The ability to do benchmarking from Rust stable
• More meaningful metrics. Maybe reliable enough to block PRs introducing regressions via CI
• gnuplot visualizations

We lose:

• The ability to benchmark private functions
• The use of #[bench]
• The community's more popular workflow.

Sample command-line output:

bench_process_transaction
                        time:   [2.4687 ms 2.4882 ms 2.5087 ms]
                        change: [-24.796% -22.249% -19.699%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  1 (1.00%) high mild
  5 (5.00%) high severe

Sample output from the HTML report:

	Lower bound	Estimate	Upper bound
Slope	2.5577 ms	2.6233 ms	2.7048 ms
R²	0.2362935	0.2435784	0.2324819
Mean	2.8168 ms	2.9047 ms	2.9977 ms
Std. Dev.	376.91 us	461.51 us	553.39 us
Median	2.6520 ms	2.8219 ms	3.0084 ms
MAD	339.46 us	489.37 us	594.31 us

bench_process_transaction.pdf

cc: #283

[mvines]

nit: "return a the transaction"