Implement the AdaGrad (Adaptive Gradient) optimizer for linear regression with MSE loss using full-batch gradient descent.
