Identify ourselves in document fetches [production]

[duckduckgrayduck]

• Adds an environment variable where we can configure our user agent when we fetch file by URL instead of it being a generic python-requests user agent which is increasingly being blocked

• Fixes a latent bug in fetch_file_url that did not retry on 500+ status codes. exc.response is falsy for status codes 400+, so it would never retry on 500. This is because the bool representation of a response with a status code of 400+ is by default False so the rest of the evaluation would become False.

>>> import requests
>>> r = requests.Response()
>>> r.status_code = 503
>>> bool(r)
False
>>> r.status_code = 200
>>> bool(r)
True

I think what this was going for is making sure that exc.response is not None to act as a guard. I added a test to make sure we do retry on 500 codes, which is the spirit of the retry logic there. When I reverted back to if (exc.Response
it failed the test.

• Adds some tests for this task because there were none to act as a regression guard.

Closes #414

Testing:
I deployed it to staging and confirmed the user agent defined in settings loads and is correctly used when fetch_url is called.