haste is a benchmarking result differ.

• Make a config file describing your benchmarks.
• Record a baseline with haste bench. This runs the benchmarks and stashes the results away in a "datum" under .haste in $CWD. The ID of the datum is printed to stdout.
• Make changes to whatever you are optimising, then run haste bench again to make a second datum.
• Run haste diff <id1> <id2> to compare the datums.

The config file is in TOML format, specified here.