Investigate model deployment

[mostafa]

Current possibilities are:

• Load the model from the filesystem accessible to GatewayD and the plugin (current method).
• Embed the model in the plugin binary.
• Create an API (in Python) that contains the model, logic and endpoints, so that the plugin acts as a client/user.

[mostafa]

The model is deployed using a dockerized TensorFlow Serving API.