Update ATLAS ingestion process

[jacobwegner]

• Remove parallel pipeline steps (for now)
• Standardize pipeline step timing / log output

This PR should speed up ingestion slightly on Heroku (the build dynos have two CPUs, which we can use to parallelize tokenization).