Docs

set to python 3.8, as 3.9 numpy under macos doesnt work as descirbed here:
https://github.com/numpy/numpy/issues/17550

the mocked function dont seem to have a __name__ attribute. However,
mocking pyfftw fails the doc creation entirely.

Fbpic modeexpansion upgrades. pyfftw is limited to 1 thread to be compatible with multiprocessing.

maintenance

remove atlas dependency
remove unsopported python2.7 (which has failed)
remove versions unsupported by conda

GitHub actions

run test in github-actions. Also allow python2 test to fail due to end of its support.

Hey @Ablinne
I have moved the fft routines to helper_fft.py to separate them better from the rest of postpic. I need them in a separate place to use them as part of the fbpic reader for the modeexpansion you suggested.
Cheers
Stephan

paricles: update calculation of beta

add new default species for QED module.

this is numerically more stable for particles with beta << 1.

by performin the fft on an aligned array

and move fft routines there which are detecting pyfftw

bugfix fbpicreader

add Fbpic reader

I have been writing a reader for the fbpic code ( https://github.com/fbpic/ ) over the last two month and I blieve that most of it is working now. There are still some convenience functions missing, that I can think of (like `Ex` and `Ey` instead instead of being forced to use `Er`, `Etheta`). But the internal mode expansion and understanding the format seem to work.

Epochreader: better simextent error msg

fix #238

related to #238

Updated image reader logic to fix #241

Fix helper.jac_det for 1d case