Currently the --dataset_definition flag specifies that we run a DAG with
a startproject job, then N workers, and an end project job.
I would like an option to insert another stage in the DAG which
would run a specified script before the endproject job.
This could, for example, use htadd to sum up histogram files
generated by the workers, etc.
Something like --extra_dag_stage=scriptname
which would run that script. I'm currently working on a standalone
script to run DAGs like this, but it would be better to build it
This can certainly wait for the client/server version.