Decouple the upload manifest from the AWS client

We currently create an upload manifest with the name 's3_upload_manifest_\d.json.\d' which contains information about the completed parts for an upload (along with the upload id). This file (and its accompanying struct) are currently (loosely) tied to S3, we should decouple them so that the upload manifest can be used for Azure.

Note that although the names differ for completed parts (and the information we save is slightly different) we can reference things is a loosely defined manner meaning we can share the implementation between cloud providers.