Better Error Handling

I think this library does a pretty good job with the "happy path", but feeding different errors into things and ensuring the errors percolate properly would probably be a useful thing to do.

The issue names no files, tests, or entry points. Start by mapping how errors move through the Scala Akka Streaming S3 client and identifying the error cases that need coverage. Done means the expected error behavior is defined and those cases propagate consistently.

Written by the indexing model from the issue text.