Ensure running 'wowutil.py build' twice works

[toolness]

Um, so when I implemented #22, I forgot to make sure that running wowutil.py build twice worked.

It looks like the WoW stuff requires the pg_trgm extension--I'm not really sure why it only complains about this the second time we try building WoW datasets though. So now we install the extension if it doesn't already exist.

It also looks like we need to get rid of at least one function that depends on wow_bldgs before we destroy it, so this does that too. Oy vey!

[toolness]

Uggh, I'm now realizing that some of the functions created by WoW also depend on NYCDB tables, which means that updating the nycdb tables by dropping them will likely cause things to explode. We may want to test updating the nycdb tables after loading WoW to make sure nothing explodes. ARGH.

[toolness]

Um, so I added a test that ensures that reloading the dependee tables doesn't raise any database errors, and they don't, so my previous comment is wrong, but I'm not really sure why it's wrong. Hmmmm.

[toolness]

Oh I think I figured it out! The get_assoc_addrs_from_bbl(text) function was preventing wow_bldgs from being destroyed because wow_bldgs was part of its function signature:

CREATE OR REPLACE FUNCTION get_assoc_addrs_from_bbl(_bbl text)
RETURNS SETOF wow_bldgs AS $$

This is unlike some of the helper functions whose function bodies query other NYCDB tables, but whose function signatures don't depend on them.

That's my working theory now, at least.