WIP: Minimal working 1D gaussian and MNIST tutorials

[NolanSmyth]

Added two simple examples/tutorials at the level of someone just learning the code

1D Gaussian with 0 mean and unit variance

• MLP architecture
• Can run on my CPU in ~ 1 min

MNIST

• NCSNpp architecture
• Can run on colab T4 GPU in ~45 minutes
• Generates samples that are recognizably numeric, but can certainly be improved.
• using preprocessing_fn to transform grayscale images from [0, 255] [0, 1] resulted in nonsense samples so I omitted this until I figure out what I'm doing wrong