Fix version of documentation depdencies

[AdamRJensen]

The versions of the documentation dependencies should be fixed to avoid the documentation breaking when new releases of those packages come out.

[jranalli]

All libraries are pinned with maximum versions in #13

[AdamRJensen]

Hi @jranalli

It was actually these packages that I was recommending to be fixed:

solarspatialtools/setup.cfg

Lines 38 to 43 in 806d047

	docs =
	sphinx
	sphinx_rtd_theme
	nbsphinx
	ipython
	nbsphinx-link

I personally would not pin the upper versions of the other packages, as they will in most cases work for later versions of those packages.

[jranalli]

Whoops, I misunderstood. I can add those in. The other packages, one of the other reviewers suggested it as a concern, because recent numpy (and maybe pandas too) updates wreaked some havoc recently. I'll try to actually evaluate which of them are a problem at the moment and try to be more judicious about it.

[AdamRJensen]

Could you make it compatible with numpy version 2 and set a lower limit on numpy?

[jranalli]

Let me look at it. I believe the issue was that the latest pandas at the time wasn't numpy 2 compatible?

[jranalli]

Yep, the issue appears to be that pandas still is using a numpy<2 dependency. So there's some environment mess there. Let me see if my tests run with that as the only pin.

[jranalli]

Alright, I'm sorry, I was misinterpreting something. There is some weird incompatibility between numpy and pandas HD5 reading that is OS and python version dependent. See for example this run that fails with numpy 2.0 and this run that passes with numpy 1.X.

My development is done on a Windows computer and so I think I was running into that issue with my own environment and had some confusion. And then I got faked out by trying to look into what dependencies pandas was specifying, which is sometimes pinned below numpy 2, and there are plenty of web results for that error message talking about incompatibility between pandas and numpy 2, possibly from right around when numpy 2 was released.

I am not adept enough with the environment management to figure out exactly what those bugs are that are affecting other versions and whether they're caused by a poor installation of pandas/numpy or something else. Now that I can see what's going on though, it appears that everything here is numpy 2 compatible as long as I limit my tests to recent versions of python (>3.10). So I think this one is closed out.

[AdamRJensen]

Beautiful