Implementing weakref and cloning

[MicahGale]

Description

This focused on making the copying experience better.

First this implemented weakref.ref for self._problem. This breaks a cyclic/recursive memory reference. This was bad for two reasons. First It made problems nearly impossible to garbage collect due to the cyclic references. Secondly it made copy.deepcopy very slow and memory costly due to the near infinite recursion in copying. To use a weakref.ref you have to call it each. So a new attribute _problem_ref was added to contain this. _problem was made a property to hide the dereferencing operation. weakref.ref can't be pickled.

This pickling conundrum was handled through __getstate__ and __setstate__. When pickling the weakref would be removed. When unpickling though all these refs were gone. There was no easy way to ensure the refs could be built at unpickling because of the order that objects are unpickled. Rather a hook was added to all contains cells, surfaces, etc. which would check if it had been unpickled and if so quickly recreate all the weakrefs.

The other issue then was that a deepcopy of a weakref still points to the same object. So the deepcopy` method was overwritten for a problem so that the references would be to the new problem, and not the old one. For all other objects when they are copied on their own they will point to the same problem as before.

All of this work sounded like clone. So that was implemented at the same time. This was done in an OpenMC like way.

Fixes #514, fixes #469.

Checklist

• I have performed a self-review of my own code
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (if applicable)

[MicahGale]

This needs more testing.

[MicahGale]

Hopefully final round of reviews.

I:

1. Implemented collection level default numbers
2. documented how this works
3. created more testing!