Add an example on how to train from a checkpoint in the documentation

In the training documentation, there is no example on how to resume training from a checkpoint. I need to add an example with --train.resume