Checkpointing approach and late publication comments

[EzequielPostan]

On the WG call of 23/Jun/2020, there was a proposal (#766) related to a notion of changing information distribution between protocol files. The change allows to improve performance for light nodes.

During the conversation, there was a comment on checkpointing in the protocol. This issue is a placeholder to discuss the feasibility of the approach given that Sidetree allows for late publication. The main questions pointed were:

1. how a late publication that represents an event previous to a checkpoint affects pruning?
2. could the state be properly updated given that information was pruned?