Multiple inserts in a row are slow because of the snapshotting

[dumbbell]

When we import definitions in RabbitMQ, we perform many #put{} in a row. Currently, Khepri's state machine asks Ra to take a snapshot every 4096 commands. With a large definitions import, this causes a lot of snapshotting.

The first suggestion from @kjnilsson was to skip snapshotting when inserting new tree nodes. Indeed, attempting a snapshot after this won't free anything: new tree nodes inserted don't replace other tree nodes. Therefore, if we don't count new tree nodes in the threshold towards the next snapshot, we might save time.