Chunk backups

As it currently is, very large backups are stored under a single compressed `tar`. This leads to unreliable uploads that can fail often. We should find a way to split our backups into multiple chunks (e.g., split the tar and then compress each one individually.)

Start by tracing how backups are packaged, compressed, and uploaded; the issue names no files, tests, or entry point. Done means defining and implementing a chunking approach that makes large backup uploads more reliable, with coverage for the resulting backup and upload behavior.

Written by the indexing model from the issue text.