[RFR] S3 bugfixes

[npow]

This PR fixes several issues which caused s3op to be stuck:

1. Need to call queue.cancel_join_thread() so that the workers can exit without flushing the queue, otherwise there is a deadlock
2. Catch the correct exceptions in download/upload (they don't actually raise ClientError)
3. Handle InternalError
4. Handle SSLError
5. Optimistically assume all other unhandled exceptions are transient

Additional improvements:

1. Added exponential backoff to jitter_sleep()
2. Set default retry config in s3op.py to match that in aws_client.py
3. Fix a bug where the retry setting was not being applied properly if config was missing
4. Fail early on fatal errors
5. Don't restart from scratch when there's no progress