Create standard benchmark interface

[lucaswoj]

This pull request defines a standard benchmark interface.

Having a consistent interface for benchmarks allows us to use the same test runner and, in the future, collect benchmark results in an automated way.

I haven't yet moved the fps benchmark over to this new interface. Once I get some 👀 and 👍 on the spec I'll tackle that.

cc @ansis @mourner @jfirebaugh

[jfirebaugh]

Looks good, I especially like capturing motivation and quality criteria in the README.

Do you think browser-based benchmarks are the way to go? Would we ever want headless / pure node benchmarks?

[lucaswoj]

Do you think browser-based benchmarks are the way to go? Would we ever want headless / pure node benchmarks?

We should explore headless and pure node benchmarks (especially #1362). The bench interface was designed to accommodate use in headless and pure node environments. For this PR, sticking to browser environments makes the most sense because it is simple and most representative of real-world usage.

[ansis]

It would be good to document how access tokens can be set and provided

Other than that, this looks good!

[lucaswoj]

One remaining task: I just realized that the env variables won't work for the debug page because it isn't run through browserify. I think it's time to combine the debug server and bench server into one.

[lucaswoj]

Alrightey. This is ready to 🚢. Since the last round of 👀 I

• combined the debug server and benchmark server into one
• allowed the use of environment access tokens for the debug server (and documented that fact)
• moved debug urls from /debug to /
• moved the existing fps benchmark over to the new interface

[lucaswoj]

fixes #2007

[mourner]

Still needs a few minor ESLint fixes.

[lucaswoj]

Oops. I could've sworn I had fixed and pushed those.

👀?

[lucaswoj]

fixes #954