Add Databricks Feature Store Component to MLOps Stacks

[vladimirk-db]

This PR adds Databricks Feature Store support as a component in MLOps Stacks. It follows similar principles applied to training code support, including sample feature engineering code, automated unit and integration testing and infra-as-code setup for deployment.

It includes the following:

• Sample feature code (transform) modules.
• Parameterized ETL notebooks that load data, apply feature modules and write toe Feature Store tables.
• Infra as code terraform modules to deploy above ETL notebooks as jobs.
• Sample model training code that is integrated with feature store client libraries.
• Automated (via Github Workflows) integration tests that trigger feature engineering pipelines and training code and post results to Github UI as PR comments.

The sample code is based on the Feature Store Taxi Example Notebook.

This functionality is optional and guarded by a cookiecutter boolean flag (include_feature_store).

Tested:

• Unit tests
• The integration test changes have been validated in a separate demo repo.
• Will instantiate and validate against a demo repo before merging.
  • Sample diff: regenerated MLOps stack on Azure w/ GH Actions with Feature Store Disabled.
  • Sample diff: regenerated MLOps stack on Azure w/ GH Actions with Feature Store Enabled.
  • In both cases all unit + integration tests pass.

TODOs (fixed):

• [x] Post-instantiation Github Actions ML code unit tests failing (when initializing a local Spark session).
• [x] Need to update docs to remove MLP references when Feature Store component is added.