Atom counts in NEB

I missed that line in the documentation (thanks!); Now that you mention
it, I actually think the current behavior is a good idea: I can envision
several scenarios in which one would NOT want to specify every single
atom, just atoms that move significantly from one side to the other.

I suspect checking for atom tags not in the initial configuration would be
an even better check than the one I proposed (and much more likely to
catch the actual problem).