Analysis of the rate of convergence of an over-parametrized convolutional neural network image classifier learned by gradient descent