This is a macro commit that squashes all commits of the development.
Here is a summary of what is done:
  - A noncollinear version of the Becke-Roussel potential to approximate the noncollinear Slater potential is implemented
  - A gamma=1 version of this is added
  - Tests are added for the exchange functional (generalized Kohn-Sham and OEP)
  - Enforcing Schwartz inequality for the square modulus of the gradient of the density (as in libxc) to improve the numerical stability.
  - Fix a problem with the kinetic energy density for k-points and spinors. The divergence now perform the basis change for non-orthogonal cell only on np points by default.
  - Adding a mixer for the vtau potential and rewriting the noncollinear MGGA for better stability.
  - Enforce that \tau > \tau^W
  - Implement the noncollinear version of the Colle-Salveti functional.
  - Fix some indices. Thanks to Fabian Tran for spotting this.
  - Adding a test for the noncollinear CS functional
  - Some cleaning of the code.