`get_feature_index_mapping` does not support `global_species`

[Luthaf]

Here is a script demonstrating the issue:

import ase

from rascal.representations import SphericalInvariants

HYPERS = {
    "soap_type": "PowerSpectrum",
    "interaction_cutoff": 4.5,
    "max_radial": 6,
    "max_angular": 6,
    "gaussian_sigma_constant": 0.3,
    "gaussian_sigma_type": "Constant",
    "cutoff_smooth_width": 0.5,
    "radial_basis": "GTO",
    "compute_gradients": False,
    "normalize": True,
    "expansion_by_species_method": "user defined",
    "global_species": [1, 6, 7, 8],
}

soap = SphericalInvariants(**HYPERS)

frame = ase.Atoms(
    symbols="CO2",
    positions=[
        (5, 5, 5),
        (5, 5, 5 - 2),
        (5, 5, 5 + 2),
    ],
    cell=[10, 10, 10],
)

X = soap.transform([frame]).get_features(soap)

mapping = soap.get_feature_index_mapping([frame])

print(len(mapping))
print(X.shape)

assert len(mapping) == X.shape[1]

This will print 756 2520 and then fail the assertion.

---

I understand why get_feature_index_mapping need to take the managers as parameter, since the features could change depending on which species are present in the structures. But get_feature_index_mapping should deal with all possible expansion_by_species_method.

---

A temporary workaround is to call get_feature_index_mapping with a specially constructed frame:

mapping = soap.get_feature_index_mapping([ase.Atoms("CHNO")])

[rosecers]

Hi! Adding a note to this issue. When masking several species, I still want them to come up in the mapping, e.g. if I only get the soap vectors for oxygen centers but there are hydrogens present, I want to see the mapping including all 1-1, 1-8, and 8-8, not just 8-8.