Rebuild for PyPy3.8 and PyPy3.9 (1.19 branch)

[h-vetinari]

Checking if we can do 1.19

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[h-vetinari]

@isuruf @jakirkham @beckermr
What do you think about adapting the pypy-migration to use numpy 1.19 (as I offered at the time)?

I don't see a reason not to TBH - this PR is passing, it would unblock the migration (there are test failures with 1.22), and it would put py{py,thon} 3.8/3.9 on the same footing everywhere numpy-wise. I'd also be willing to do the backports for 1.20 & 1.21, so we can keep bumping incrementally according to NEP29.

[mattip]

I don't know if there is a demand for NumPy + PyPy versions other than 1.22, but if it is "cheap" to do so it might be nice.

[h-vetinari]

Planning to merge this soon-ish. Whether the migration gets adapted is a separate question, but otherwise this has no downsides I can think of.

[isuruf]

Agree with @mattip. If you have a use case for this, let's merge, but if not lets close this PR

[h-vetinari]

The "usecase" I'm thinking of is to have less variation in numpy host versions for the packages we build. CPython (e.g 3.10) is different because there, old numpy versions are simply not amenable for backporting, but numpy 1.19+ is source-compatible with 3.8/3.9 and so I think we should do it for pypy, given that it's very little effort (which I'm willing to do, at that).

[mattip]

Thanks!