Investigate deleting job artifacts proactively

[farski]

When a Porter job creates an intermediary file (such as the source file being ingested, or a transcoded file before it's copied to its destination), those files are created in the artifacts S3 bucket. Those files are only used in the context of the Step Function execution, and unneeded after the job stops running, other than for debugging purposes.

Currently, cleanup of those files is handled by a lifecycle rule on the bucket: objects expire after 1 day.

We do a large enough volume of work in Porter that at any given time, that bucket is over 200 GB. So we're currently paying for every object to live for 24 hours, when they are only providing value for about a minute, in most cases (under 10 minutes in nearly all cases).

We could have the state machine proactively delete objects associated with the job right before it completes to significantly reduce our byte-seconds usage in S3.