Closes #3329: hstack to match numpy

[1RyanK]

Since hstack uses concatenate in numpy, I redid concatenate on server side. It seems to work with normal pdarrays but not Strings or Categorical. I made some modifications on other things that also rely on ConcatenateMsg.chpl.

I also updated vstack because I somewhat doubt that functioned properly before. I noticed that the code where the vstack and delete tests were (tests/array_manipulation_tests.py) was not included in the pytest.ini, so I added it. The tests for delete failed, but I didn't touch it, so I suspect that was ongoing prior to my adding it.

While trying to clean things up a bit, I noticed some interesting behavior. Some things rely on the concatenate with ordered=False (meaning that when concatenating two arrays, the result may have the two arrays mixed together, rather than a whole block where you have the first, followed by everything in the second). Specifically, when comparing two Categorical objects together (i.e. for equality), it would concatenate the Strings of categories and also two aranges (which were, I guess, meant to associate the categories to numbers). Basically, it seemed to rely on the fact that they would get concatenated in the same way. This seems very sketchy to me. At some point we should probably further investigate set_categories in arkouda/categorical.py and the functions it calls. It takes a shockingly long time to test equality on two Categoricals with three things, but different categories. Definitely feels like optimizations are possible. Also ConcatenateMsg.chpl could probably benefit from optimization for Strings.

Closes #3329: hstack to match numpy

[1RyanK]

I keep getting arkouda/pdarraysetops.py:427: error: Cannot determine type of "shape" [has-type] approximately on the line shape1 = arrays[i - 1].shape. Wherever the first reference is to .shape mypy decides that it doesn't know what type that is, and that that's a problem. However, there is no problem in hstack's version. I tried printing out some info about the shape and got this:

(3,)
<class 'tuple'>
<class 'int'>

Note that I printed out the shape, type(shape), and then type(shape[0]).
I would greatly appreciate help resolving this and then after I add some tests for hstack and vstack it should be all set to be merged in.
Another thing I wasn't sure about came from here: #2802
Should I set max_bits to the minimum or maximum of the positive max_bits across the input arrays? In the example there's only one. I currently have it set to minimum, which would throw away bits. Kind of like in significant figures you don't pretend you have precision that you don't actually have.