Cloud restore could use the Rift indexes to minimise the amount of data streamed from the cloud

When doing a 'FullScan' from the Rift index we are given the start/end sequence numbers that we would like from this vBucket. The (local) binary data writer currently just opens the data store on a restore and discards the data it doesn't need from the buffered writer.

We could be smarter about this (for cloud) by using the 'GetObjectRange' function to get an 'io.Reader' for only the data which we are going to restore.

We could do the same for local restore, except we simply use 'Seek' to skip the first 'n' bytes.

I doubt this will make a massive difference to the costing for restore since we will be reading almost all the data from the data stores anyway but it's definitely worth doing.