Fix SOC DOS parsing for latest pymatgen

[kavanase]

Since materialsproject/pymatgen#4239, Spin.down entries are no longer stored in pymatgen DOS objects parsed from SOC calculations (as desired). There was a patch in sumo to delete this entry for SOC calculations if present, to make later downstream plotting easier, but this now causes a KeyError when used with the latest pymatgen due to this entry no longer being present:

sumo-dosplot
Traceback (most recent call last):
  File "/home/mmm0650/miniconda3/bin/sumo-dosplot", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/home/mmm0650/miniconda3/lib/python3.12/site-packages/sumo/cli/dosplot.py", line 646, in main
    dosplot(
  File "/home/mmm0650/miniconda3/lib/python3.12/site-packages/sumo/cli/dosplot.py", line 207, in dosplot
    dos, pdos = load_dos(
                ^^^^^^^^^
  File "/home/mmm0650/miniconda3/lib/python3.12/site-packages/sumo/electronic_structure/dos.py", line 132, in load_dos
    del dos.densities[Spin.down]
        ~~~~~~~~~~~~~^^^^^^^^^^^
KeyError: <Spin.down: -1>

This small update fixes the issue.

[ajjackson]

Would it be feasible to produce a small test case for this, or are these vaspruns always monstrous?

[kavanase]

I looked at doing this when I added the fix, but didn't see a test file for DOS loading/handling. I've added one now. I checked and found the Cs2SnBr6 example in the repo is a SOC run so works for the test.

[ajjackson]

That's wonderful, thanks Seán! Please add a line to the CHANGELOG and I'll be happy to merge.

PS I notice you are merging from a branch named "master". You may find that if we squash this PR or accept another one first then git pull stops working and you are stuck out of sync. It's salvageable, but generally easier to work on a branch with a new name instead...

[kavanase]

Done.

Yeah should've made a new branch first, but went with the quick option. Will try remember that next time!