Optimization of EPW, mainly in the evaluation of e-ph vertex on coarse grids

See merge request !802

Phonon mag noncollinear qe 6.5

See merge request !780

Update COUPLE examples for new laxlib

See merge request !803

My first optimization of EPW, mainly in the evaluation of e-ph vertex on coarse grids

See merge request epw/q-e!3

Update EPW/src/elph2.f90, EPW/src/elphel2_shuffle.f90, EPW/src/elphon_shuffle.f90, EPW/src/elphon_shuffle_wrap.f90, EPW/src/epw_init.f90, EPW/src/io_epw.f90, EPW/src/kfold.f90, EPW/src/low_lvl.f90, EPW/src/rotate.f90 files

Develop

See merge request !801

Revert the symmetry call as previously.

See merge request !800

The symmetry matrix need to be created in all cases as it might be
needed in the case of restarts.

Cp optimization

See merge request !799

Remove trailing white spaces in EPW

See merge request !798

Added line under IF (imq == 0) THEN

See merge request !792

Atomic charge and magnetic moment estimate in MD and relax runs

See merge request !797

with atoms fixed at their original positions. This made such estimate 100%
useless (instead of 99% in the absence of such bug). Noticed by Vic Bermudez.

One more fix for CUDA mp interfaces

See merge request !796

Non-parallel file systems with pools

See merge request !793

Fix GPU implementation of mp_root_sum

See merge request !794

file system with k-point parallelization. Notice that each pool will write
wavefunctions to a different scratch directory.  In order to avoid this and
collect all wavefunctions into a single directory, much complexity must be
added to the writing code. I don't think this is worth the effort.

IF (noncolin) CALL errore("elphon_shuffle_wrap", 'TR temporarily disabled with SOC, please use noinv=.true. in pw.x calc preceding ph.x calc', 1)

Tells user to not rely in inversion symmetry in ph.x calculations 

Update the soc test in EPW following the SU2 commit.

See merge request !791

Su2 rotations for spinors

See merge request !790