Efficient recovery of CommitLogStoreActor by saving a snapshot

[xirc]

Motivation

CommitLogStoreActor behaves as an event source actor.
Saving a snapshot makes a recovery process efficient if the number of events increases.

Design

The team will address this issue by the following.
Note that some of the details might be changed.

• Require a snapshot store always
  • Provide a migration guide
• Use lerna.akka.entityreplication.raft.eventsourced.persistence.snapshot-store.plugin to configure a snapshot store plugin ID. The default value is empty string "".
• Save a snapshot after a certain number of events.
  • We can configure the number by using lerna.akka.entityreplication.raft.eventsourced.persistence.snapshot-every.
  • The default value is 1000.
• Add at least two public APIs to RaftSettings.
  • def eventSourcedSnapshotStorePluginId: String
  • def eventSourcedSnapshotEvery: Int
• Add at least one public API to ClusterReplicationSettings (both classic and typed)
  • def withEventSourcedSnapshotStorePluginId: ClusterReplicationSettings
• Introduce a new state class and save it as a snapshot
  • This class contains only a value of typeLogEntryIndex now
• Add a unit test of CommitLogStoreActor
• Update documents (/docs/*)