Add training progress indicator to SGD optimize_batch

[selfint]

not needed, progress indicator will be in a separate (to be added) train function, similar to the progress indicator in mnist example train loop.