Fix a bug in AtomsBase interface

[tjjarvinen]

This fixes a bug in AtomsBase interface that caused extra vector data, like forces, to not load into Atoms structure.

To see the bug you can use this as an input

using ExtXYZ

text = """2
Lattice="2.614036117884091 0.0 0.0 0.0 2.6528336296738044 0.0 0.0 0.0 3.8250280122051756" Properties=species:S:1:pos:R:3:force:R:3:tags:I:1 config_type=FLD_TiAl spacegroup="P 1" virial="5.072173561696366 0.1220123768779895 0.6518229755809941 0.1220123768779895 4.667636799854875 0.5969893898844183 0.6518229755809941 0.5969893898844183 4.700422750506493" energy=-1703.64063822 unit_cell=conventional n_minim_iter=2 pbc="T T T"
Ti       1.30924260       1.32316179       1.62637131       0.86219000       0.78737000       2.65969000        0
Al       0.11095015       0.09471147      -0.05013464      -0.86219000      -0.78737000      -2.65969000        1
"""

fname=tempname()
open(fname, "w") do io
    print(io, text)
end

atoms = ExtXYZ.load(fname)
@show atoms[1][:force]

I also fixed writing AtomsBase data vectors, like force, and added tests for both cases.

As a context of this bug. This peace is the final part to allow training ACE models using AtomsBase interface.

cc @cortner

[jameskermode]

Thank you for fixing this. I've tested locally and in CI and am happy to merge. I will then tag a new release.