Somehow preserve S3 metadata across calls

[EdSchouten]

The awssyncer will inherently need to do a full synchronisation every couple of hours. The reason for this is that inotify may easily miss events in our case. For example, in case files are hardlinked, we'll only receive a single event for one of the paths.

Still, there is room for improving the number of requests we send over to S3. If we would cache the state of S3 across calls, we'd know exactly what to sync and what can be skipped.