here the coul/long part is split off into a separate
pair style. that should allow better optimization for
the very long coulomb cutoff. it also is a base test
for adding new methods of long-range coulumb treatment.
through using of a longer realspace cutoff for the
coulomb interactions and the (new) pppm/cg kspace
style that loops only over charged particles there
is a massive speedup of the calculation close to 2.5x.
