the previous  z(:,i) = CMPLX( z1(1,:,i) ,z1(2,:,i) ) avoids getting error message "3-rd index is 2 but upper bound
is 1" when doing array-bound-checking    

...Christoph asked for this because a too low precision breaks the respective requirements in Spex.

Fixed small bugs to allow to compile without SCALACK and parallel HDF5, added configuration for QuantumMobile image

Initialize flo in io/writeBasis.F90 to overcome valgrind complaint that it may have uninitialized entries

...not relevant at all.

necessary to keep the eig-file, because otherwise the phase-relationship
between Bloch functions and Wannier-functions is not known. Therefore, in the
past, the eig.hdf file was generated during a first run, where the mmn and amn
files are generated. In subsequent runs of the wannier-branch the eig.hdf file
could be reused by setting eig66=T in the inp file. This ensured consistent
phase relations between bloch functions and wannier functions and therefore
a consistent choice of phases between all files generated in the wannier branch.
Since the eig66 switch has disappeared in the MaX-Release, I try to revive it
with this commit. 

...at the moment this is buggy. I will fix it in the next commits.

...also some cleanup in init/initParallelProcesses.F90.

Valgrind complained that the respective variable might not have been initialized.