add automated performance (timing) benchmarking

[atravitz]

following up from conversation in PR #1858, it would be good to have an automated way to quickly verify that we haven't introduced some huge performance regression.

[IAlibay]

Generally I agree. However, following the discussion on that PR I disagree that this should be "one more thing we shove into devtools".

The best way to do this is either to clean up performance_benchmarks or to create a proper ASV performance benchmark. The latter is minimal investment if you start with a couple of small entries.