Update mini-batch-gradient-descent-from-scratch.ipynb

[RajKulk16]

Change in coef_der

Old -> coef_der = -2 * np.dot((y_train[idx] - y_hat),X_train[idx])

New -> coef_der = -2/self.batch_size * np.dot((y_train[idx] - y_hat),X_train[idx])