Since we do not store which executions produced binary data, for pruning
on S3 we need to query for binary data items for each execution in order
to delete them. To minimize requests to S3, allow the user to skip
pruning requests when setting TTL at bucket level.