Initial Draft of APE16: CPU Threading - #46

FYI, in case you want to continue the APE: I asked for Scipy, and they don't use OpenMP because it's not considered portable enough. They use very little multi-core overall: mostly in the linear algebra packages they call into (don't know what those use), and very few functions use Python multiprocessing from the Python standard library, and even there because of issues on some platforms they are considering using something like https://github.com/joblib/joblib that's apparently more robust. I do think a short APE that is a summary of available options in Python and what other projects use would be very useful, as a basis for discussions what (if anything) to do in Astropy to improve performance in the future. Not sure if an APE would be the appropriate format for that, it could also be a wiki page, or a poster / proceeding at e.g. a Scipy conference "Options for multi-core CPU processing for scientific Python packages in 2019".

@jamienoss - are you still interested in pursuing this APE? If you don't think you have the time/desire, it may be we need to consider this APE withdrawn. We can then close this PR, but then you can leave your branch in place so that if anyone wants to they can pick up this topic (but include you as a co-author of course if they start from that text!)