Switching to wyckoff surface atom tagging

[brookwander]

Use cases:

1. This will be used implicitly by anyone using catlas to enumerate slabs. It makes it so the undercoordination of possible surface atoms is directly compared to its bulk counterpart (via wyckoff) rather than considering the min coordination for that species in the bulk. This removes the need to consider surface atoms by height as well, which could have resulted in subsurface placement in the past.

Scope:
What will be done:

1. The Surface class from ocdata.surfaces will be brought into catlas and modified.
2. It will primarily use pymatgen.structure.Structure objects rather than ase.Atoms
3. Atom tiling will be modified to be handled on pymatgen.structure.Structure objects rather than ase.Atoms
4. atom coordination will be considered by full_wyckoff rather than min per species
5. _find_surface_atoms_by_height will be removed
6. Some modifications to catlas.enumerate_slabs_adslabs.enumerate_slabs to support above changes
7. Possibly some integration with the broken bond model to reduce compute although how to do this isn't immediately obvious because BB model as implemented does not use Voronoi nn
8. testing: look at the differences between adslabs enumerated (up to max miller =2)by the previous approach and this current one for 300 randomly selected bulks to quantify the impact of this approach. For a subset of differences (say 200 slabs) ensure the change is in fact an improvement. Although I don't know why it wouldn't be. Maybe try out a couple of NN approaches and compare. Voronoi is slow and has some problems. Take a look to see if this completely corrects issues with large adsorbates. Test overlap post adslab enumerations.
9. doc updates

What will not be done:

1. Full integration of existing ocdata infra. This has been something we have talked about and I think we are close to that now, but it will not be completed here.

Execution plan:
This will be a solo project all to be completed by Brook. The execution plan is well detailed in the scope section.

Consider any problems / risk:
Unforeseen issues with the new tagging method. There should be very little risk once the testing plan has been completed.