Update dump-to-s3 step function to "unparallelize" entire thing

[shawncrawley]

All of the dump to s3 logic (i.e. the pg_dump command) occurs on the same exact machine, which will quickly get strapped for resources when dealing with the size of some of the dumps. Thus, we need to update the step function to synchronize the processing of all of the dumps - namely, ensuring that the "egis" dumps and "viz" dumps are not parallel processed, and same with processing schemas versus tables. Right now the step function branches out in to essentially four tracks, but it should really just be a single track.

[RobHanna-NOAA]

Note: If we are able to get larger EC2's for processing, which is expected as the current "bastion" server is far, far too small. TBD