Restart system comparison check doesn't account for GCMC waters

[lohedges]

The check for system equivalence when performing a restart currently doesn't account for the fact that the checkpoint system will contain additional waters used for GCMC insertions. This needs to be accounted for when comparing the number of molecules.

[lohedges]

In addition, it looks like the GCMCSampler isn't being reconstructed correctly for restarts.

[lohedges]

Ah, it does for repex, but not for the regular runner since the repex state isn't stored as part of the checkpoint. I'll think of a way to add this.