Pin Hail version(s) used in data pipeline

[mattsolo1]

We should pin the latest version of Hail used to successfully run data pipeline tasks.

Could be in specified in data-pipeline/requirements.txt.

However since some pipeline tasks are computationally intensive and may be infrequently run, the last successful Hail version could differ across tasks, so perhaps Hail version could be stored on a per-task basis.

See discussion: https://github.com/broadinstitute/gnomad-browser/pull/922