Machine learning with the mnist data set

I tried doing this mostly in safe rust, which resulted in horrendous code, don't do this.

Running

python src/mnist.py to get training data
cargo run --release <epochs> <batch_size> <learning_rate> <image_range>

Where:
epochs is the number of times to train,
batch_size the number of times to split example images,
learning_rate is how quickly the model "learns" (lower == more accurate),
image_range is start and end index of images to test from, for example: 1-5 will test images from 1 to 5 from the mnist data.

Followed Magicalbat's video