Pull style Anti-Entropy

[bbengfort]

Implement pull anti-entropy with lamport scaler timestamp and replica version ids.

• Each replica writes versions to it’s own local log via a monotonic increasing version number that implements a Lamport scalar timestamp. E.g. object versions are (replica version, replica id).
• The replica version is incremented for each local write, as well as set to max(replica version, neighbor version) during each anti-entropy session.
• Anti-entropy sessions reconcile the per-node logs of each replica by appending writes to the logs that are behind for each participant. These sessions can be push/pull/bilateral.
• Each replica keeps an object: latest version mapping and always writes to the latest version, no matter which log that version happens to be in.

[bbengfort]

We've decided to go to a single log for all replicas in the simulation.