Compare how different batch sizes affect gradient descent on linear regression, and return a list of loss curves (one per batch size).
