Merge spin-orbit code and enhance it

Ramón Cuadrado has a working version already, but need to streamline the handling of the new dimensioning arguments (h_spin_dim, e_spin_dim, spinor_dim, etc), which get in the way of other
merges. A first proof-of-concept will involve the "diagon" hierarchy.

This code is extremely big and changes a lot of the routines in the siesta code base.
The merge has revealed a substantial amount of inconsistencies regarding spin names/dimensions. This is one of the major hurdles regarding the future development of siesta.

Currently this branch has been tested against the following tests:
si001 si2x1h-dipole si64 sic-slab si-optical wannier h2o fe_clust_noncollinear born born_spin

Work items:
Push spin-orbit branch into lp:~siesta-spinorbit-developers/siesta/trunk-so: DONE
Notify TDDFT-devs about MERGE: DONE
Notify LDA+U-devs about MERGE: DONE
DOCS Hamiltonian elements (NC/SO): INPROGRESS
DOCS Density matrix elements (NC/SO): INPROGRESS
DOCS Energy density matrix elements (NC/SO): INPROGRESS
DOCS Grid elements (NC/SO): INPROGRESS
Check atom.F for inconsistencies (I cannot find the error message per Alberto) : POSTPONED
Change "Magnetization" to "Spin": DONE
TESTS of NC/SO: DONE
TESTS of NP/SP: DONE
Check dhscf and other routines for possible bugs regarding the spin components: DONE
Document code on where NC/SO does not work: INPROGRESS
Document utilities that are NOT NC/SO compatible: INPROGRESS
Parallel MULLIKEN: DONE