Feature request: Docs: add section on how to chose batch size

[JRMeyer]

OOM errors happen all the time in training for newcomers and pros... we should have a simple guide on using --reverse_{train,test,dev} and steps to choose the right batchsize on some setup...

also, maybe we just need --reverse_batches instead of all three --reverse_{train,test,dev}... I can't think of a situation where you'd want a subset, but not all, set to reverse