Snapshot Replication

Make snapshots a full object which can be replicated between similar storage backends. Definable states would also be documented.

Replication of snapshots can be the most efficient way to implement both volume migration and volume mirroring. Beyond that the same mechanism can be used to build additional features:

* Archiving of snapshots on additional servers to allow a volume to be "restored" by cloning it
   at a new location. Unlike mirroring there is no need to explicitly pair active/passive servers
   in advance. Snapshot replication can achieve disaster recovery schemes at a lower cost
   than mirroring (albeit with a longer RPO).
* Replication of snapshots can also be used to distirbute master images to poential host
   machines, where a new boot volume can be cloned from the replicated snapshot with
   far less bandwidth being required at provisioning time. This can enable more rapid
   deployment of compute nodes from master images, and allow for updates of master
   images to be less costly.