feat: benchmark stream web readable

[RafaelGSS]

I have tried to create a fair comparisson between these implementations.

However, I'm not sure if the results are confidents. I really would like to get your thoughts here: @mcollina @jasnell

[RafaelGSS]

Results I got locally:

streams.Readable reading 1e3 * "some data" x 2,465 ops/sec ±0.71% (87 runs sampled)
streams.web.Readable reading 1e3 * "some data" x 471 ops/sec ±0.39% (86 runs sampled)
Fastest is streams.Readable reading 1e3 * "some data"

[jasnell]

The results are not surprising. When running just a single iteration of the web streams vs node.js streams side-by-side, the performance comparison between the two should be nearly identical. However, when running it over-and-over, the web streams will be much slower for a number of reasons -- but mostly because of the heavy reliance on promises and the additional object allocations it requires. There are likely a number of performance optimizations that could be made internally but the web streams API was not designed for performance as much as it was designed for portability.

[RafaelGSS]

I see. Actually, I came across this benchmark because I'm investigating the undici.fetch performance (nodejs/undici#1203), and following my analysis, one of the slowest paths are in the pullAlgorithm that makes use of WebStreams (Readable/Writable).

The benchmarks in undici are also running it over-and-over, so the results make sense to this benchmark.

Thank you!