AWS upload/download implementation does not respect the applied manifest
The upload/download storage method in the AWS implementation stores object as path/dataset-name without respecting the applied "GENERIC" manifest that is expecting block to be numbered from 0 to N. sdutil upload object as single blocks so these must be named as "0" or consumer application won't be able to read them back (according with the generic manifest)
Change object_name = f"{s3_folder_name}/{dataset.name}"
to object_name = f"{s3_folder_name}/0"
in both upload and download method