1072 qc automate state properties in qualitycontrol and evaluation based on metric states

[dbirman]

This PR refactors the overall_status and evaluation_status fields to built from the status of the QCMetric objects in each evaluation. The complication to this PR is that it would require people to build QualityControl objects to use these features, if you dump to JSON and you don't re-validate as a model you can only access the private status lists indirectly. You can still use the schema without building the classes you just don't get access to the convenience functions that are defined using the @property tags.

Obviously the hope is that nobody has to deal with this stuff since the QC portal handles it for them.

I'll walk through the tests to explain how this works:

        test_eval = QCEvaluation(
            evaluation_name="Drift map",
            evaluation_modality=Modality.ECEPHYS,
            evaluation_stage=Stage.PROCESSING,
            qc_metrics=[
                QCMetric(
                    name="Multiple values example",
                    value={"stuff": "in_a_dict"},
                    metric_status_history=[
                        QCStatus(evaluator="Bob", timestamp=datetime.fromisoformat("2020-10-10"), status=Status.PASS)
                    ],
                ),
                QCMetric(
                    name="Drift map pass/fail",
                    value=False,
                    description="Manual evaluation of whether the drift map looks good",
                    reference="s3://some-data-somewhere",
                    metric_status_history=[
                        QCStatus(evaluator="Bob", timestamp=datetime.fromisoformat("2020-10-10"), status=Status.PASS)
                    ],
                ),
            ],
        )


        q = QualityControl(
            evaluations=[test_eval, test_eval],
        )

This object has status set for the metrics, but not for the evaluation or the full quality control object. If you now call:

        q.evaluate_status()

You'll find that self.assertEqual(q.overall_status.status, Status.PASS)

If you append a metric to one of the evaluations that is set to Status.PENDING or FAIL the evaluation status and the overall status will subsequently resolve to that state:

        # Add a pending metric to the first evaluation
        q.evaluations[0].qc_metrics.append(
            QCMetric(
                name="Drift map pass/fail",
                value=False,
                description="Manual evaluation of whether the drift map looks good",
                reference="s3://some-data-somewhere",
                metric_status_history=[
                    QCStatus(
                        evaluator="Automated", timestamp=datetime.fromisoformat("2020-10-10"), status=Status.PENDING
                    )
                ],
            )
        )

        q.evaluate_status()
        self.assertEqual(q.overall_status.status, Status.PENDING)

Timestamps automatically get set to datetime.now() unless the user specifies a timestamp, to allow for generating the qc.json separately from the actual processing that happened.

[saskiad]

let's use the same datetime that we use throughout the schema.

I'm still a little confused by the structure overall - I might have you explain it for me again (or look with fresh eyes in the morning)

[dbirman]

This PR now also closes two other issues that involved adding asset_id fields to QCEvaluation and adding a validator to QCMetric to make sure that a QCStatus object is included when pushed (and not just an empty list in metric_status_history)

[saskiad]

I want to deal with the asset id separately. Otherwise I think it looks good.