Enable slaves to switch masters when masters are temporarily unavailable

[krishan1390]

Due to #72 and after changes in #73 , the following scenario breaks

1. • Start 3 masters in raft quorum with steps mentioned in Service discovery changes to push and get status updates of master and slaves #76
2. • Start 1 slave. Auto Service discovery enables it to replicate from any master
3. • Do some writes
4. • Shut down the master that the slave is replicating from. Can get the master from - /bin/dkvctl -dkvAddr 127.0.0.1:8001 -getClusterInfo <"dc id">
5. • Wait for slave to replicate from another master
6. • Do some more writes
7. • Shut down the new master that the slave is replicating from. Can get the master from - /bin/dkvctl -dkvAddr 127.0.0.1:8001 -getClusterInfo <"dc id">
8. • Do steps 6 and 7 until slave starts replicating from a previous master again. Once it gets to this stage, replication will fail due to slave having higher change number that master

Specifically because the change numbers aren't consistent across raft nodes (when one of them restarts it applies all changes to its own state), this trickles to slaves too which can thus have a higher change number than another node and is thus now not sure what change number to replicate from