Key space/depth operations commonly used in SISR models are missing

[Artoriuz]

Hi! I would like to ask if it would be possible to add support for the depth to space and space to depth operations commonly found on many SISR models.

While you can achieve the same result with some clever usage of keras.ops.reshape and keras.ops.transpose, both TF and Pytorch offer them as easy to use functions:

https://www.tensorflow.org/api_docs/python/tf/nn/depth_to_space
https://www.tensorflow.org/api_docs/python/tf/nn/space_to_depth

https://pytorch.org/docs/stable/generated/torch.nn.PixelShuffle.html
https://pytorch.org/docs/stable/generated/torch.nn.PixelUnshuffle.html

I've been able to find the same operations in the PIX library as well:

https://dm-pix.readthedocs.io/en/latest/api.html#depth-and-space-transformations

This exact same depth to space transformation can be found in the following example:

https://keras.io/examples/vision/super_resolution_sub_pixel/

Thanks in advance =)