We improved our timeout and retry logic when uploading API batch job operations and results to S3.

We use S3 to store operations and results for API batch jobs. We weren’t handling timeouts when communicating with S3 as well as we could have; these fixes will improve reliability and throughput of our API batch jobs.