Further updating POTCAR validation / identification

[esoteric-ephemera]

Summary

In some releases of VASP POTCARs, there exist multiple POTCARs with the same TITEL. The previous summary statistics used to uniquely ID POTCARs relied on the uniqueness of TITELs. The current implementation allows for multiple POTCARs within a single release to have the same TITEL field, but different data. The summary statistics have also been updated.

For context: those with a PSP license, take a look at the PBE pseudopotentials W_pv and W_pv_new. Both have the same TITEL, thus W_pv_new was likely intended to be an in-place update of W_pv. However both are "valid" POTCARs.

Some other things to note for future POTCAR releases:

• Some POTCARs are misplaced, e.g., the PBE C_d POTCAR is an LDA POTCAR, but officially released in the PBE library. That requires changing how the validator searches for possible matches
• Some POTCARs were updated in place by VASP without version control, e.g., Eu and other lanthanides in the PBE library. Thus multiple Eu POTCARs exist, with different TITELs, and are all valid

[janosh]

Just to leave a paper trail: this project of overhauling the pymatgen POTCAR validation to move away from hashes and towards checking several aggregations of numerical data in a POTCAR against reference values started in #3351, was updated in #3370 to support validating the latest 6.4 POTCARs and received this bug fix to make sure identically titled but numerically different POTCARs (of which VASP published ~20 in the original PBE POTCARs) all check out as valid.

We hope this is now productions ready. User feedback welcome!