Add mechanism to indicate if objects were dropped and a fullsync is needed

Several new stats added:

rt_source_errors - non-persistent
rt_sink_errors   - non-persistent (sink errors detected on the *source* side)
rt_dirty         - persistent across restarted, cleared by a successful fullsync

When a rt_dirty > 0, the leader will add the node to the list: fullsync_suggested

When rt_dirty is > 0 during a fullsync already in progress, the node will be added to the list: fullsync_suggested_during_fs This list will be moved to fullsync_suggested when the currently running fullsync completes.

Fullsync replication is not using enough nodes

Fixed in build

Full-sync replication using more than max_fssource_node

Fixed in build

All RT source nodes connect to same sink when bouncing the source cluster

(randomized IP addresses) shipped in previous build, will be included in current build Fixed in build Real-time replication is automatically going to configure itself when new nodes are added.

Replication migration during shutdown dropping objects

We've fixed what we can for now, the other fixes require 1.3.