Extend artifact handling for an ML Model type

[keithmanville]

Add a new artifact type to store models. The model type (matching to available types in ML Flow) must be specified as a part of this artifact type.

See here for an example call to log a keras model.
https://mlflow.org/docs/latest/python_api/mlflow.keras.html#mlflow.keras.save.log_model

This feature should work with the following MLFlow model types:

• keras
• scikit learn
• pytorch
• tensorflow

This should otherwise leverage that the artifacts extension to the task engine yaml uses.

Blocked by:

• As a user I would like to be able to store output from a single task plugin or from a combination of task plugins as artifacts for later use and analysis. #655

Definition of Done

• Model artifacts can be stored as part of the outputs of a job
• The model types defined above all work.
• Extend the artifacts unit tests to include the model artifact type
• The MNIST example should be updated to use this feature and still work
• The feature is merged into dev.