A backup replica should be switch off not only with disconnected backup primary (task:
INDY-1681) but with backup instance degradation. It shouldn't send BACKUP_INSTANCE_FAULTY but should collect its own messages with this type.
- Implement a new strategy (see
INDY-1681) which will check for performance degradation (the same way as master does) in addition to disconnections
- switch off a backup replica when it has more then M its BACKUP_INSTANCE_FAULTY messages in K time.
- add tests to check correct replica switch off
- testing performance changes (shouldn't be worse) after disconnect or degraded a backup primary node.
- testing memory consumption (should be better in case a backup primary node degraded)