Generate model hashes and populate experiment_models table when generating train tasks

[ecsalomon]

For beginning evaluation or checking results prior to an experiment finishing, it is helpful to have the experiment_models lookup. An additional nice-to-have would be a model_status table or add status to experiment_models. This should play nicely with #440.

[thcrock]

So by status, what exactly do you mean? A switch between "not yet trained", "trained", and "skipped"?

[ecsalomon]

@saleiro or @avishekrk came up with the status, so I'll let them answer :)