qexsd_init_total_energy was internally converting energies to Ha from Ry.

In CP, this was leading to printing of (harmless but incorrect) half ot the true value of all energy terms. I think that interface routines in qexsd.f90 should accepot Hartree both when reading and when writing. NOTE: the call to qexsd_init_total_energy in qexsd_step_addstep was in my opinion not correct before (now I think it is)