Remove old runs and pipelines

[donkirkby]

Once all the old pipelines are converted to child containers in issue #751, we can remove all the code to support the old pipelines:

• runs
• fleet and sandbox
• docker and slurmlib
• pipelines, methods, transformations, and code resources
• compound datatypes
• delete orphaned datasets from MD5 conflicts
• delete orphaned datasets from intermediate steps that weren't wired to pipeline outputs
• add tests for removal plan
• remove @types/jquery from package.json Can't remember why we wanted this, so postpone it for now.

Run the migrations and the find_orphans management command as part of deployment.

[donkirkby]

What should we do about intermediate datasets that were not wired to pipeline outputs? They got recorded as datasets in the old runs, but they would not get created in the converted container runs.

Should we just delete the datasets? What about the MD5 conflict datasets? Should we delete those, too?

[rhliang]

I think that's OK, considering they would have gotten purged in pretty short order otherwise anyway.

[dmacmillan]

I wrote a django management script to handle orphaned datasets, which should clear the last two things off this list. The last thing to do would now be to update the test server where kive is running with the drop_pipeline branch. I think the test server is called bulbasaur but I still don't have access to it.

[donkirkby]

Closed by 79617ae.