Problem: we don't have a way to run unit tests in CI and locally when writing code for the inference-services repo.
For example, to run tests locally we need to:

• create a venv
• install dependencies (pytest etc..)
• run pytest via PYTHONPATH=$(pwd):$PYTHONPATH pytest test/unit

Ilias started some work in https://gerrit.wikimedia.org/r/c/machinelearning/liftwing/inference-services/+/982396
Ideally we should have a way to run tox and get everything set up automatically.