XDCR - P2P - bg process to pull ckpts under error conditions

XDCR P2P will perform a checkpoint pull from all other nodes prior to pipeline startup just in case that the node itself is newly rebalanced in and peers haven't had a chance to push ckpts over.

In the case of a network partition, this won't be possible. Or, if timeout occurs in a very slow running system.

Instead, XDCR should start the pipeline, and launch a bg process to keep polling until network partition is repaired. Then, if there are newer checkpoints that has been pulled and can be used (i.e. no failover occurred), then restart pipeline