(Idea) feature: set optimizer from outside

[YuriCat]

example

net = SimpleConv2dModel2()
optim = Adam(net.parameters(), lr=1e-4)

learner = Learner(args=args, net=net, optim=optim, remote=False)
learner.run()

This PR requires #170 (return model instance by calling net())

[ikki407]

I like this idea because the optimizer often affects the performance.