Does batch normalization replace 'layers.experimental.preprocessing.Rescaling' in CNN models?

No. Rescaling and batch normalization are different concepts.

You rescale your input data by a fix scale and offset to be in a certain range. So it is a preprocessing step. Furthermore, it does not have trainable values. Often 0 to 1 or -1 to 1 are used.

In normalization, you want to scale your feature maps in such a way, that you have zero mean with unit variance. Batch normalization normalizes the batch at a certain stage within the model, e.g. after a convolution. It's parameters are trainable, usualy denated as gamma and beta. With these parameters, the model can scale the feature, since probably zero mean/unit variance is not the optimal scaling to converge your training objective.