Add docker-compose setup and make onboarding easier

[toolness]

This makes it easier to develop autoapi with docker-compose, and also fixes a few things:

• Fixes #58.
• Uses COPY instead of ADD.
• Optimizes use of the build cache when generating layers for the image. Previously, changing any files in the source directory would rebuild the entire image from scratch, which takes quite some time and means bigger downloads from docker hub; now layers are reused so that a minor change to the source code results in a very quick image rebuild, and a top-level fs layer change so very little needs to be downloaded from docker hub.
• Adds a docker-compose.yml to make it easier to develop (not just deploy) the site via docker-compose up.
• Re-organizes the documentation a bit to make it easier to understand and get started with things.

To do:

• [x] Document the new functionality, making sure to distinguish between deploying with docker vs. developing with docker.
• [x] Document Docker's existing use of the data_sources directory. This took me a little while to figure out.
• [x] Propagate environment variables into the docker container, either through docker-compose.yml or something like python-dotenv.

[toolness]

Sorry this does so many things in one PR... Ideally I would've put them in separate ones, but since they all built on each other, it seemed most expedient to just do it as one.

[toolness]

Oh, I should also note that there are lots of warnings logged when I start the server--most of them appear to be because we're using APIs that have been deprecated in the latest versions of the libraries we're using, which we're retrieving because most of the lines in our requirements.txt use >=. I'd like to address that in a separate PR though, since this one is already rather big as-is!

[catherinedevlin]

Thank you so much for this great PR, @toolness !