This implements a custom docker image that the pipeline should run in on Cloud Dataflow. It contains all the necessary python/linux dependencies for Variant Transforms.
Notable impacts:
VT now uses this container by default. You can override which container (documented), but you cant avoid using one.
I've moved everything to python 3.8 (from 3.7) as part of this work
updated cloudbuild to include building this 2nd container
This implements a custom docker image that the pipeline should run in on Cloud Dataflow. It contains all the necessary python/linux dependencies for Variant Transforms.
Notable impacts:
fixes #705