Make s3 syncing more efficient

[moredip]

Currently I think we do a network call per file in order to get the etag so we can see whether the file needs updating or not. We can probably do 1 call to get the metadata for all files in a bucket, and then use the results for all the local files.