Load and run a trained model in the labeling UI

[gabriel-v]

Add a --eval_model argument to labels.py. When this argument is provided, the script will load that model and print its prediction for each image pair displayed by the UI.