Collections: DCP system events should transmit name and uid of the affected collection

A short coming has been identified with the current DCP system event specification, it doesn't include the name of the collection (or the scope.name)

The issue is that some clients, such as backup may need to recover the collection config, e.g a restore of a backup to an empty bucket will need to create collections before the documents can be written.

If the backup service can only get collection-UIDs from DCP system events and DCP mutations, then windows of inconsistency can occur if the backup service is required to find out collection names by looking at say ns_server's view of collections. Maybe ns_server has dropped the collection, yet the backup needs the name.

It is a much simpler system if the DCP events can include the scope and collection names, we should really aim to have the ability for DCP system events to transmit enough data that a DCP client could rebuild the entire JSON manifest.