Why we need to change these QE routines:

beta_mod.f90 we deallocate tab_beta only if it has been allocated.
             Moreover we need to use tab_beta so we make it PUBLIC.

bfgs_module.f90  If the bfgs routine exceedes the maximum number of 
            allowed iterations there is no way to change the 
            bfgs_file in a different pw.x run. Added line 1147.

bp_mod.f90 In QE Berry phase works only without pools. This change is
           needed to make it work with both pools and G parallelization in
           thermo_pw.

clocks_handler.f90 we increase the defaults number of clocks.
                   At line 45 maxclocks=2048

lmdif.f90 we use at line 95 epsdiff=1.d-6 instead of zero.

io_dyn_mat.f90 The length of the atm vector became 6, but this routine was
             not changed

parameters.f90  we increase the maximum number of k points
                At line 16 npk=120000

phq_init.f90 A bug for GPU + 1k point per pool + ph.x corrected.

pwcom.f90 igk_k deallocated from the acc system only if allocated.
          igk_d_d and ngk_d allocated again. They are needed inside tpw.

rhoat_mod.f90 we need to call deallocate_tab_rhoat so we make this routine
              public at line 25

vloc_mod.f90 we deallocate tab_vloc only if it has been allocated.
             See at line 238.
