Fe calculations within the GGA.

Two pseudopotentials. One of them
has a small p core, and looks ugly.
Increasing r_c(p) fixes this, but..
is the pseudopotential more or less
transferable?

The "spin" files are experimental.
