Clean up MyriaSysTemp tables

[jortiz16]

Users have reported that MyriaSysTemp tables might stay in Postgres in cases when there is a failed (or canceled?) ingest. Depending on the size of the relation, these tables take up room and there is no direct way to delete or access them.

We need to figure out an automated way to clean up these tables and decide when to do it. @mbalazin suggested calling this clean up operation as a query (through use of the DbExecute operator). That way, the user knows this clean up happened.

If we want to do this automatically, here are some options:

1. At a Myria restart
2. After a worker recovers

Any others?

Tagging: @shrjain @senderista @jingjingwang

[senderista]

For now, how about checking for and deleting these tables during worker initialization? Then a global garbage collection is as simple as a restart of the Myria service.