Support MIC and properly normalise site vector

[GabrielBram]

Implement support for a site with neighbours that stretch across periodic boundaries. Also fixed non-normalised vector for lone pairs > 1.

[robinsonmt1]

Adds in check for minimum image when finding vectors to each neighbour atom, so should now work for atoms on the edge of a unit cell.

The placement and rotation should also still work even if the adsorbate is outside of the unit cell (previously still worked when the adsorbate was big enough to stretch outside of the unit cell even if the adsorption site was necessarily inside the cell)

[robinsonmt1]

Causes failure of the CI tests. Without this line, neighbour lists and site_norms are consistent between cutoff_mult values. When this line is included, the neighbour lists are inconsistent, with cutoff_mult=1 missing the Si atom (atom index: 2) and cutoff_mult=1.2 producing a different site_norm with a length of 0.9999999963586863 (inc. norm) vs 1.0000000034771082 (not inc. norm).

[robinsonmt1]

The site_norm difference is fine, because of the precision of positions in ASE up to 1e-8. The error comes from the ~6e-9 difference, giving positions with a total error of only 0.276. This will just require changing the computed positions to the new ones with the normalised site_norm

[robinsonmt1]

Currently I can't work out why the neighbours_list gives what it does for the 1th shell in the MeOH placing (shell around the O atom), even without using the cutoff_mult.

Ideally, after the H is added, the MeOH should be added directly above the H (O-H-O = 180[deg]). The following is the behaviour and stats for cutoff_mult values from 0.9 - 1.1:
(Distances: O-Al = 1.61 A, O-Si = 1.61 A)
cutoff_mult = 0.9: correct behaviour, only includes atom 0 in the 1th shell (cutoffs: O = 0.594, Al = 1.089, Si = 0.999)
cutoff_mult = 1: incorrect behaviour, includes atom 1 (Al) in the 1th shell, giving an incorrect site_norm vector (cutoffs: O = 0.66, Al = 1.21, Si = 1.11)
cutoff_mult = 1.1: slightly incorrect behaviour, includes both atoms 1 (Al) and 2 (Si) in the 1th shell, partially cancelling out and giving a near-correct site, but with an O-H-O angle of 170.5[deg]. (cutoffs: O = 0.726, Al = 1.331, Si = 1.221)

My issue with these is that if the neighbours function does use the covalent radii/natural_cutoffs, then it seems like the Si atom should also be included in the cutoff_mult=1 and the Al atom should still be included even in cutoff_mult=0.9. I think I must be missing or miscalculating something.