Migrate group_mean_continuity_check validation tests to dbt

[aesharpe]

Summary

Toplevel validation migration spreadsheet for reference.

There are four group mean continuity checks we want to migrate into dbt:

table	python source
_core_eia923__cooling_system_information	pudl/transform/eia923.py:cooling_system_information_continuity()
_core_eia923__fgd_operation_maintenance	pudl/transform/eia923.py:fgd_continuity_check()
_core_eia860__fgd_equipment	pudl/transform/eia860.py:fgd_equipment_continuity()
_core_eia860__cooling_equipment	pudl/transform/eia860.py:cooling_equipment_continuity()

All four ultimately call validate.py:group_mean_continuity_check().

Current status

There is a draft implementation in #4092, which includes a demo instance of the test for _core_eia923__cooling_system_information. The maximum acceptable number of outliers for this instance of the test is 2.

• In dbt, the test fails with 4 outliers.
• In dagster, this test succeeds, finding only 2 outliers.
• In a python notebook pointed at the output parquet files, the test fails, finding the same 4 outliers dbt finds.

Next steps

Figure out why dagster doesn't find the other two outliers but a python notebook can.

[aesharpe]

Paused investigation on this in order to get the rest of the tests migrated

[krivard]

Confirmed that:

• this is still happening with an otherwise-working Dagster env
• because they are currently implemented as Dagster asset checks, they're not blocking removal of pudl/test/validate