The only difference between the unit and integration docker-compose config files is that the integration version has a timescale writer container that runs for integration tests.

The setup and code that is run is the same, so it makes sense to have a single set of containers for running both tests.

We can integrate both of these files together. We should not start the timescale writer when the unit tests are run, but we can run it manually when the integration tests are needed. It would be good to be able to reuse the existing code in test.sh for bring up a test instance without running tests (test.sh -u and test.sh -d) for the integration tests too, to make the development cycle faster when running these tests.

One way to separate the integration and unit tests is to use pytest marks, like we have in BU: https://github.com/metabrainz/brainzutils-python/pull/44#discussion_r520778420. This avoids the need to hard-code the path to the integration tests when running them.