WIP: source-free exchange-correlation magnetic fields

This MR adds two options to pw.x:

• the Reduced Stoner Theory discussed in PRB 86, 064437 (2012), controlled by the input parameter ssxc (Scale Spin part of eXchange Correlation), and
• source free B_{xc} proposed in J. Chem. Theory Comput. 2018, 14, 3, 1247–1253, controlled by the no_source option. This generally requires the ssxc option introduced in the previous point (see paper).

This is still a WIP for many reasons:

1. I barely started testing the code. I only have a few results for BCC Fe that I will (hopefully soon) post here.
2. Way too many useless FFTs are performed in remove_Bxc_source. Data should stay in G space all the time.
3. The subroutine remove_Bxc_source is polluted with a bunch of code I used for testing. I think it can be useful for developers outside the QE community to get familiar with the data structure used in the code, but it must be removed from where it is. Or maybe this stuff is already documented somewhere and (unfortunately!) I missed it.
4. B_xc is computed twice when no_source is active and the changes to v_xc are meant to preserve as much code as possible, but are clearly sub-optimal from a logical and computational point of view.
5. Maybe no_source is not the best name for this option 😕

I'm posting the MR hoping that someone out there will have the time to review the changes and share her/his opinions. Maybe exceptionally kind souls may also help with testing! 👼