Scaling Large Data Science Environments with Spark and Kubernetes