Add max/mean unpooling operations

[AzamatB]

These are fairly common operations in image2image models (including 3d images) where they go in hand with ConvTranspose layers.

[MikeInnes]

I'm on board; I think it'd make sense to add them to NNlib. If there are any volunteers I'd be happy to look at a PR.

[bhvieira]

What is meant by max/mean-unpooling? Is it meant like in the U-Net where you return the max value to the position it was found in the descending path? I would like something like that, but I'm not sure we can do that without binding the downscaling and upscaling pooling layers.

[darsnack]

Anymore clarity on this? My interpretation is that it is just #1468 in which case we can close.

[DhairyaLGandhi]

No unpooling is slightly different. Check the pytorch layer with the same name

[kraftpunk97]

Has this issue been addressed? Can I claim this?