Convert image classification and generative examples to use X10 by default

[BradLarson]

With the release of the 0.9 toolchain containing X10, we now have a higher-performance alternative to the default eager execution mode. In the interest of having swift-models demonstrate best practices, we should have our image classification and generative models use the highest performance pathway available. For image classification and image generative models, that's pretty much always X10.

Therefore, we should make the (small) modifications needed for the following models to use X10 by default:

• [ ] Custom-CIFAR10
• [x] ResNet-CIFAR10
• [x] LeNet-MNIST
• [x] VGG-Imagewoof
• [x] MobileNetV1-Imagenette
• [x] MobileNetV2-Imagenette
• [ ] GAN
• [ ] CycleGan
• [ ] DCGAN
• [ ] FastStyleTransferDemo
• [ ] pix2pix

This is most likely a follow-on issue to issue #511, because the conversion of training loops to Epochs will conflict with updates to the training loops here. We also may need to check that X10's CPU compile times and execution speeds are acceptable for macOS.