Separate training command

[BMJHayward]

As an end-user of VSpark, I'd like to train models a separate command/workflow. Currently I can only save models from the importance command, which is convenient, but I don't always want to run an importance analysis.