uninitialized submesh in LDA+U

There is currently a bug in X(orbitalset_get_coefficients):

  use_submesh = os%submesh
  ! Because of possible phase corrections at the border, the array X(orb) is always stored
  ! on the submesh for complex wavefunctions
#ifdef R_TCOMPLEX
  if (.not. has_phase) use_submesh = .true.
#endif

This means that later on we refer to the submesh even if they are not defined. It is not clear at the moment, whether we should simply replace that by #ifdef R_TCOMPLEX if (.not. has_phase) use_submesh = .true. .and. os%submesh #endif

or whether the code has to be changed such as to ensure that submeshes are initialized in the complex case without a phase.