Model validation for FS models

[aliazzzdat]

FS models cannot predict as pyfunc models yet (issue #70), thus we can't use MLflow Evaluate with FS models.

However, MLflow Evaluate can take a function instead of a pyfunc model_uri. This means we can call an FS model in the function and use this function in the Evaluate method.

Although this is a good solution, we still cannot enable baseline comparison, as we need to provide a model_uri of a pyfunc model (so this means it's not possible to compare a new FS model with a baseline FS model).

[arpitjasa-db]

@aliazzzdat thanks for all the changes! Looks like some tests are failing, can we fix those and also do a full instantiation of Stacks to confirm everything is still working as expected? Version changes can be tricky, and want to make sure we don't break anything

[aliazzzdat]

Featurization pipeline works on DBR 15.3 and with MLflow 2.11

Model training pipeline works on DBR 15.3 and with MLflow 2.11

Validation works with a Feature Store model

[aliazzzdat]

The batch inference works as expected once we create the feature_store_inference_input table