Allow distance routines to accept an NumPy Array or AtomGroup [Core]

[hmacdope]

Fixes #3708
Successor to #3718

Second attempt at addressing #3708 using a python only design. CZI_EOSS4 performance track.

Changes made in this Pull Request:

• Change @check_coords to take also AtomGroups and return positions
• Atomgroup parsing in @check_coords must be enabled with argument allow_atomgroup with default False maintaining current behaviour
• Added type hinting to lots of distances module
• Change distance_array, self_distance_array, calc_bonds, calc_angles, calc_dihedrals, and apply_PBC to take an atomgroup
• Added tests for AtomGroup input to distances. Still some code duplication here can be possibly cleaned up .

Do we want to also change nsgrid functions, and the transform_ functions? If so is that best done here or separate PR?

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

[IAlibay]

Not 100% sure why some of this isn't being picked up by pep8speaks, anyways just a couple of PEP8 changes for the lines that changed (I'm ignoring one of the pep8speaks one because 81 char >> readable than 79 char in that occasion).

[hmacdope]

Thanks @IAlibay much appreciated.

[hmacdope]

Hmmmm still getting what looks to me like NumPy versioning issues on the Azure pipelines despite #3737 merged?

[IAlibay]

How odd, was developed merged into this branch?

[hmacdope]

Yeah it looks all up to date AFAIK. Log shows #3737 as merged in my history.

[IAlibay]

@hmacdope - Here we go this is the problem: https://github.com/numpy/numpy/releases/tag/v1.21.0

"A runtime-subcriptable alias has been added for ndarray"

Essentially NDArray isn't going to work in 1.20... That's a bit of a pain, I guess our options here are to either vendor the alias somewhere in our code or break NEP29. We should probably move this to a new issue since it's a pretty big decision.

[orbeckst]

That's just related to typing, though, right?

It looks to me that typing will remain a moving target for some time so my view is to do the best you can within current constraints and then move on. Add a comment how to do it nicely when np 1.21 is lay of the land and then put in something ugly or less specific.

I recognize that there are different views and mine is just the one of an outsider as far as typing goes...

[IAlibay]

That's just related to typing, though, right?

It looks to me that typing will remain a moving target for some time so my view is to do the best you can within current constraints and then move on. Add a comment how to do it nicely when np 1.21 is lay of the land and then put in something ugly or less specific.

I recognize that there are different views and mine is just the one of an outsider as far as typing goes...

Hugo will open an issue later today, but one of the things we'll need to discuss is what we do with the outreachy typing project since I think they've already started using npt.NDArray

[orbeckst]

Hugo will open an issue later today, but one of the things we'll need to discuss is what we do with the outreachy typing project since I think they've already started using npt.NDArray

Ahhh... and that's not been caught in tests with lowest supported versions? Alright, that's discussion that should happen on its own issue.

[hmacdope]

Pending discussion on #3748

[hmacdope]

Other than the flaky tests this should be good to go again following discussion in #3748.

[hmacdope]

@MDAnalysis/coredevs any chance I could get a second review here? It would be good to get this and #3751 merged soonish. :)

[IAlibay]

@hmacdope can you address whatever can be addressed from pep8speaks (all seems reasonable stuff like two lines vs 1). And get CI to run green. I'll review tomorrow once that's done.

[IAlibay]

@hmacdope here are all the suggested changes that were left.

I'll eagerly approve with the hope that you'll handle these suggestions before merging this PR.

[hmacdope]

Thanks @IAlibay, sorry I swear I made the changes before I pinged but I think I confused myself with #3751.