Currently the deposit loader reads the deposit artifacts from the deposit server via an internal api.
That api streams the archive to the loader without naming it.
Forcing the deposit loader to hard-code a name (currently configured through a "default_filename" constructor attribute in the loader).

This is currently consistent so no harm done.

A more suitable approach would be to make the deposit server api send the archive name
alongside the streamed archive (through http header for example).

That changes would need a second change in the deposit loader as well so that the loader
reads such information to name the retrieved archive appropriately (and then drop the
"default_filename" from the loader constructor).