Investigate refactoring the flagging script to a dbt incremental Python model

[jeancochrane]

In theory, the flagging script should be able to run as an incremental and Python model using dbt. In practice, we don't have any experience using incremental models yet, and there are some signs that the Python integration may not be fully baked yet (see the known issues for dbt-athena Python models). I think it's worth taking a day to spike a refactor that uses an incremental Python model to determine if it's a promising direction.