DOS normalization (FHIaims, VASP, exciting)

This MR resolves #728 (closed) and #868 (closed). It is accompanied by https://github.com/nomad-coe/electronic-parsers/pull/16 on the parser side.

Basically, the DOS in FHIaims, VASP, exciting were incorrectly scaled at the parser side: they no longer lead to integer electrons when integrating over a band. Moreover, the total DOS in older FHIaims versions was off by a factor of 2 (regardless of spin). Projected DOS or bands were not affected. This MR updates the parsers and adds the new metadata entry raw_data_normalization_factor, to store the original FHIaims DOS without any correction. Up to now, this metadata only applies to FHIaims, so I could also move it to that parser's specific metadata.