[W.I.P.] Update benchmarks: Drop ujson, add hyperjson, add orjson (issue #130)

[hartwork]

For #130

Should not be merged right now because:

• in my current results (see updated benchmark tables), stdlib json is 2-3 times faster than rapidjson, not slower. That's unexpected — any ideas what I might be doing wrong? What do you get when you run make benchmarks-tables?
• depending on progress at https://github.com/ultrajson/ultrajson-fork/issues/5#issuecomment-590002891 we may want to (a) re-add ujson soon, (b) later or (c) not drop it in the first place

@lelit what do you think?

[lelit]

Hi!

I tried to replicate the problem with ujson, but it works for me (Debian sid here), either with Python 3.7 or Python 3.8... I thus opted to keep it around.

I also did not notice the difference between RJ and standard library json.

OOC, on which OS/hardware did you run the benchmarks?

Anyway, thanks a lot for your initial implementation and research.

[hartwork]

I tried to replicate the problem with ujson, but it works for me (Debian sid here), either with Python 3.7 or Python 3.8... I thus opted to keep it around.

The underlying link issue was fixed upstream now at ultrajson/ultrajson#312 . I'm not sure why import does not fail with a link error for you, too.

I expect ujson 2.0.0 to perform rather differently than 1.35 so it might be worth benchmarking against (both ujson 1.35 and) 2.0.0 later.

I also did not notice the difference between RJ and standard library json.

OOC, on which OS/hardware did you run the benchmarks?

16 GB RAM, Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz, 2 cores, 2 threads per core, Gentoo Linux

Closing PR becauce of:

• 275eeda
• a5ec68f
• 616b91c