Implement vanilla gradient descent and momentum-based gradient descent to minimize the MSE loss of a linear model, and return both loss curves.
