feat(scripts): vacuum tables after import

[philbooth]

Related to #20 (comment), adds VACUUM commands for each table after data is added to them.

Strictly speaking, in most cases we only need SORT ONLY rather than FULL. However the scripts do DELETE rows if force_reload is set, and the redshift docs state that the saving for SORT ONLY isn't normally huge, so it seems prudent to opt for FULL.

I timed some VACUUM commands earlier and they took about five seconds to complete. I didn't check the stats beforehand though to see whether they had much to do, so that may not be a fair indication of how the import scripts will be affected in practice.

You'll notice the weird extra END; that each VACUUM block starts with. This is the only way I found to work around the error VACUUM cannot run inside a transaction block. I also tried SET AUTOCOMMIT, which didn't work.

@rfk, r?

[rfk]

I'm +1 to this, and I think the vacuum should complete in a timely manner if we do it regularly.

This is the only way I found to work around the error VACUUM cannot run inside a transaction block

This flips my weirdness sensor though, I'm going to take a look and see if I can understand what's going on here before merge.

[rfk]

OK, looks like the "postgres" wrapper module runs everything in a transaction by default, and there's not an easy way to disable that. We might try moving to the lower-level "psycopg2" module but r+ on this for now.