Bugfix for Vasprun reading with POTCAR for chemical shift calculations

[MichaelWolloch]

Summary

Major changes:

• Fixes a bug updating the charges from POTCAR while reading a vasprun.xml file in the case of a chemical shift calculation
• Slight update to the splitting of POTCARs when reading from file.
• Added a test for the chemical shift vasprun.xml parsing with a POTCAR
• Re-added a test for internal POTCAR hashes.

This work is a result of updating atomate to work with the newest pymatgen and custodian releases and figuring out a hard to spot bug that was present there since May. Please see https://github.com/hackingmaterials/atomate/pull/778#issuecomment-1662336368 for details.

Another discussion related to this was taking place in #3068. There the hashing system was (hotly) debated and the initial problem was fixed (afaik). I still think that this PR has a nicer version, which is easy to read and preserves the internal hashes. This is why I re-enabled that unit-test. However, I am fine with commenting that out again or use the new PMG_POTCAR_CHECKS environmental variable to make it conditional?