DOS metainfo layout

Pavel brought up some good points about the inflexiblity and inconsistency in how atom projected, species projected and total dos are presented in the metainfo. @ladinesa also discussed this in one of the metainfo refactor workshops.

• section_atom_projected_dos and section_species_projected_dos are not designed in a way that each atom/species could have a different (l,m)-projection. Pavel says that at least WIEN2k and Lobster would need this functionality.
• section_dos seems to also store (l,m) projections. These are, however, not for the total DOS, but seem to be atom projected by looking at the shape. This is inconsistent with the metainfo description. At least the VASP parser seems to fill the atom projected DOS here.
• dos_values has no unit, should be 1/energy?

We should agree on a new schema for storing this information.