Benchmarking

[quinn-dougherty]

We would, at some point, like intelligence about

• relative speed of the different strategies/modes (Flag to change the method of distribution add/subtract/multiply/etc #326 / Issue326 - asMode type  #379)
• whether or not parse trees grow worse than linear in cost, or at what size does a squiggle string become prohibitively expensive.

Some related ideas, but aren't strictly speaking benchmarking:

• what is the failure rate of property tests
• for a given test, what is the maximum precision at which it passes?

After #386, we might want benchmarking be it's own package ultimately.