Unit tests for Python backend code

[isms]

Go through and add unit tests for Django/DRF or Python functions in the prediction service. (This is correctness testing separate from ML evaluation.)

Tests should not unduly slow down the build.

---

Points will be awarded continuously through the end of the competition -- this issue will not close.

[WGierke]

Would it be useful to add integrations such as CodeCov to continuously monitor whether a test really adds more coverage / a feature reduces the coverage?

[isms]

Good thought, we have used and enjoyed codecov.io.

Maybe we can start with just a report in the test pipeline, and staff can look into setting up the integration if time permits.

[WGierke]

Does something like #282 make sense? It generates a coverage report locally which could then also be persisted by Travis (like this one).