+ let it generate the correct triangle of the matrix.

+ add some debug output in hybrid branch

...to make the mysterious segmentation error in the allocation of the
former facfac array magically disappear. Now it is a stack array. I did
not identify the cause for the error.

...related to the choice whether the nonlocal potential should be added in a certain iteration

Changed the defaults in atom_input.f and added some timers&stops to prevent this in future.