Mission: baseline model showdown — step 4 of 7

Complete the trained rung: sweep every threshold and keep the one that maximizes accuracy ON TRAIN — a real trained model, one parameter big.