One attempt to fix ChemEnv error in tests

[JaGeo]

Summary

As discussed in #2788, one way of fixing the ChemEnv error in the tests would be converting the PeriodicNeighbor object in a PeriodicSite object. I have implemented this. It might however affect the execution speed.

If someone else, manages to fix the as_dict from PeriodicNeighbor in a cheaper way, I would prefer this, of course.

[JaGeo]

Linting still fails but tests are running through.
Please let me know what you think and if you have any better/faster ideas (@shyuep , @janosh). Both species and coordinates are influenced when using from_dict and as_dict from PeriodicNeighbors.

[coveralls]

Coverage: 78.018% (+7.9%) from 70.155% when pulling f31ab03 on JaGeo:bug_fix into 811dd75 on materialsproject:master.

[shyuep]

Thanks. This is fine. I'd rather it be working and slower. Someone else can go figure out a faster method if it bothers them. Most people would not care.

[janosh]

This doesn't look like much a slowdown anyway.

[janosh]

Thanks @JaGeo. Good to have green CI again. Like you said, makes contributing much easier. I'll close #2788 now.

[JaGeo]

Just adding this here:

pymatgen/pymatgen/core/structure.py

Line 158 in c1a7f5d

def coords(self) -> np.ndarray: # type: ignore

This name is a bit unfortunate. If as_dict as inherited from PeriodicSites is then used, "coords" (typically fractional coordinates but now cartesian ones due to the overloading) is used and this might lead to the other problems with from_dict and as_dict for PeriodicNeighbors (besides the dict issue for the species).

[JaGeo]

In any case, I am happy that the issue is fixed now. Thanks @shyuep @janosh