Rely on application to run database migrations for docker-compose

[jszwedko]

Curious to get thoughts on this one!

I personally view it as a responsibility of the application to run database migrations. Rather than build a separate docker image that is the mdillon/postgis + the dumped schema from rake db:migrate and friends (imported in the entrypoint), we can instead use the image as-is and rely on docker-compose to run the migrations as part of docker-compose up.

Overall I think this reduces maintenance burden, but I'm open to other thoughts.

Depends on #284

[bigfleet]

@jszwedko I am 💯 in favor of this getting a flavor of citygram/data image, with a specific tag. I do think these data images represent the opportunity to pull representative datasets (for example, to let some dev somewhere deal with what happens when we DO have e.g. 500k rows for a city over a time period). But your use case above is much more common and just as needed. That use case is overall in good position to use some tag-- feel free to nominate it and build support for it into any pipeline that makes sense. (There might not be anything to put into the image at that point!)