Resigned the internal structural object atom_loop() method to be up to 20 times faster.
The base model_loop() method is no longer used at all. It was previously being used in a nested
fashion, making the loop extremely slow when lots of models exist.