Add basic benchmarking using criterion

[emwalker]

This PR adds a simple benchmarking harness as an example of the kind of thing that can be done using a library called criterion.

The criterion library is a well-known Rust library that is often used for this kind of thing. The benchmark in this PR is just an example; hopefully it will be straightforward to extend to more substantial cases. Its comparisons from run to run are a little noisy, but I suspect this can be dialed in. Examples of people using criterion here and here.

To run the suite, do:

$ make bench
$ cargo bench

I'll leave this PR in draft since I didn't discuss it first.

[jaytaph]

Awesome. Looking forward to merge

[emwalker]

This PR should be ready to review and merge now. The PR is just a placeholder. Feel free to make significant changes, or refactor, or even revert the functionality if it doesn't prove useful.

I adjusted the settings in an attempt to make the results a little more consistent from one run to another. The lesson is that you have to run cargo bench several times in order to get a picture of where things are at.

[jaytaph]

@emwalker Thanks for the PR.

Is this something we should run on each PR and deploy the results? (we have the *.developer.gosub.io domain for this for now)?

[emwalker]

I doubt the benchmark results are deterministic enough to block a PR on at this point. But I can imagine in the future failing a PR check if some time threshold is passed. I suspect the actual benchmark might need to be modified if it was going to be used for a PR check. So perhaps good to leave for a future investigation.