Support pickle model import on API server

[santi1234567]

Motivation

Using blockprint currently involves having the training data in a folder. Once the model is trained, it can be persisted, but imported back into the application.
This feature would allow anyone to use already trained models, as well as saving a lot of training time and disk space.

Description

It is needed to create a function that can import the model back, and call this from the main application.
Depending on whether the model was given or not, we shall import it.

TODOs:

• Create function to import model
• Create environment variable to allow user to input a model path
• Depending on the environment variable, load the model at execution time

[michaelsproul]

We should use a non-zero exit code for errors

Suggested change

	exit(0)
	exit(1)

[michaelsproul]

I'm going to delete this because it conflicts with the intention of DISABLE_CLASSIFIER. Sometimes it is useful to start the API server without a classifier, e.g. if you just want to serve info from the sqlite DB.

If we want to simplify this in future and always require a classifier, we should remove the DISABLE_CLASSIFIER option

Suggested change

	if classifier is None:
	print("The classifier was not loaded")
	exit(0)

[michaelsproul]

Merging as-is and will push my changes in a follow-up PR