Each time a new file was loaded, all its atoms would be appended to all other presently loaded
structures. This would cause large slow downs of the atom_loop() function, and a few other
problems. Now structures are referenced by structure index rather than model number.
