Improve proof compression via diffs rather than snapshots

[James-Oswald]

Currently proof (and AEG file) compression does not exist, we just use the entire JSON representation of the tree at each step for proofs. A better option is to do what git does and represent each inference rule application in the proof file by only listing the changes made at each step (what nodes were created? where?). This would significantly reduce proof size.

• Create a JSON schema for the new proof file format
• Implementation of saving proofs as just the set of changes
• Implementation for constructing proofs from a set of changes