More circular statistics, and use those for Distributions of Longitude (and Angle?)

[mhvk]

What is the problem this feature will solve?

In discussions on fixing Distribution so that LatitudeDistribution and LongitudeDistribution can be created, @eerovaher noted that the standard pdf_mean, pdf_std, etc., are not really useful for Longitude, given the wrapping. We probably should use those in stats.circstat instead (or at least give the option to), but those do not have median, mad, and various sortings (including histograms) are missing as well (and are obviously tricky!).

Describe the desired outcome

Ideally we implement a full complement of circular statistics. In #14421 (comment), @nstarman linked to an introduction of circular statistics that includes a median, https://insidebigdata.com/2021/02/12/circular-statistics-in-python-an-intuitive-intro/

Additional context

No response

[mhvk]

cc @larrybradley and @mirca

[pllim]

There is also this issue:

• Circular Statistics: astropy vs scipy #13749

[nstarman]

I agree #13749 is an ongoing issue! And it would be nice to upstream. However, I don't think this PR exacerbates issue 13749 as it adds statistics which should appear in any stats library, so we would need to make sure we and scipy have median stats, regardless.