Add functionality to compare parsers to the reference implementation

[grahamgower]

Parser implementations wish to check that their output matches the reference implementation (e.g. in demes-c https://github.com/grahamgower/demes-c/blob/main/tests/compare-reference-implementation.py). To avoid duplicating that functionality into demes-python (popsim-consortium/demes-python#430), we should probably include this in the spec repo.

Note that this is distinct from the graph equality checks that we have in demes-python, which ignores various fields such as the metadata, description, and the current time units (it converts to generations).