Stypox

The other day I noticed that SkyTube uses NewPipeExtractor and they have [a few commits](https://github.com/TeamNewPipe/NewPipeExtractor/compare/dev...SkyTubeTeam:NewPipeExtractor:dev) on top of our `dev` specifically for retaining compatibility with API 19+. In NewPipe we...

Mmmh but then you would lose Java's requirement/guarantee to put `throws ...` annotations on methods, and we would therefore be more likely to forget that a method is fallible, because...

In addition we could put the checkstyle check as a git push hook, provided it runs quickly enough. The Rust compiler repo has something like that, and I have to...

> I really don't like git push hooks because when I edit a Readme or update mocks I don't care about checkstyle rules. It's quite easy to skip the hooks...

Ok you convinced me :-)

After making sure nanojson produces JDK 11-compatible binaries, the CI here builds (and all tests pass).

I collected 10 samples from before and after, and I calculated the average and standard deviation. Given that data, we can't conclude which of the two commits is faster, since...

Wait I wrote this simple performance test that just tries to open all mock JSON files. It takes 1.8s on 837705a1a4751236a3dd31e68b3e2613e786ae86 (before) and takes 26.5s on 1f6cb35de98238c2c95078a4d2b8da74c02e5d49 (i.e. this PR)....

@FireMasterK I already discovered why: that set of JSON files contains very long strings and the reusableBuffer's size was growing linearly instead of exponentially

@FireMasterK yes that works. I actually used the diff below but it's doing the same thing; I found that 1.3 also works fine as a constant which might save some...