Testing improvement: Add integration tests with controlled data

[jcpitre]

Describe the problem

In #448, it was proposed to add new test framework that would consists of integration tests using controlled data and an actual database.

Proposed solution

The DB would be filled with:

• A reduced version of source.csv that we control
• The equivalent of test_datasets.json to add missing data that can reproduce error conditions

Note that you can refer to populate_db_test_data.py for code that fills a DB from a json file.

The test DB should be hosted either on a local Docker container, that would let us run it from the command line, or on GCP or Github when run from a Github action.

Alternatives you've considered

No response

Additional context

No response

[davidgamez]

Tasks:

• [ ] Create all test cases in the JSON file and document them
• [ ] Unify loading of test data. Load data only from a JSON file
• [ ] Start tests with clean local DB and delete all entities after the complete test. Use the docker liquibase-test DB.