v4: insert_batch should fail if you try to run it with something already in signal_load

[krivard]

We recently repaired the mergedkey branch of v4 to force it to run dbjobs automatically after insert_batch. This is because the fix_islatest query, having been stripped of the process_status column, performs quite poorly when you wait to run dbjobs until all the CSVs have been run through insert_batch. Each additional CSV file added to signal_load increases the number of rows participating in the fix_islatest join, but a smaller and smaller percentage of rows are updated.

We want to:

• Add a check to insert_batch so that if signal_load has anything in it, it throws an exception
• Add a test (could be an integration test if you did the database setup, or a unit test if you mocked out the database responses) to confirm that insert_batch throws an exception if you run it with something in signal_load

[melange396]

i have a call to logger.fatal here which i somehow boneheadedly thought would actually halt execution. this needs to raise an exception at the very least. it looks like @xavier-xia-99 might have gotten this done in #948 but i cant make sense of the mass of changes in there... this should be included in v4 before it goes live, so ill get to adding the exception shortly.

[krivard]

Fixed in #967