test: add benchmark

[ctavan]

Add a simple benchmark to be able to catch order-of-magnitude
performance regressions.

[ctavan]

It does not yet have browser support but I believe that if at all, uuid performance might be critical only on the server (think of a system that generates events at a very high rate and uses uuid as event IDs).

I have no proof, but I would assume that such high-frequency use cases for uuid are much less likely in the browser.

[awwit]

@ctavan I think performance testing in the browser should also occur locally. If we delegate this task to a third-party service, then distortions in the results may occur. Since we cannot know what tasks are performed in parallel on third-party machines…

[ctavan]

Adding browser support was easy enough so I just added it:

[ctavan]

Agree @awwit. I think if at all, this benchmark will only allow us to compare before/after of a certain change on the same machine.

I don't think we can get comparable results over time and I also don't intend to do this.

[broofa]

FWIW, here's the results for node 12.10.0 on my laptop (same laptop I used for the other results I posted)

uuidv1() x 1,472,274 ops/sec ±0.62% (91 runs sampled)
uuidv1() fill existing array x 4,389,392 ops/sec ±0.44% (87 runs sampled)
uuidv4() x 348,408 ops/sec ±0.75% (87 runs sampled)
uuidv4() fill existing array x 390,717 ops/sec ±1.32% (84 runs sampled)
uuidv3() DNS x 137,150 ops/sec ±0.49% (88 runs sampled)
uuidv3() URL x 126,044 ops/sec ±0.75% (89 runs sampled)
uuidv3() MY_NAMESPACE x 129,816 ops/sec ±0.45% (92 runs sampled)
uuidv5() DNS x 135,630 ops/sec ±0.68% (88 runs sampled)
uuidv5() URL x 123,146 ops/sec ±1.03% (90 runs sampled)
uuidv5() MY_NAMESPACE x 127,863 ops/sec ±0.67% (84 runs sampled)
Fastest is uuidv1() fill existing array

[broofa]

A few suggestions, otherwise looks good. Thanks!

[broofa]

This isn't of interest to most readers. Move to PR template (preferably the hidden .github/pull_request_template.md file).