MaxPool: correct pooling description (#86559)
In the documentation of `nn.MaxPool2d` and `nn.MaxPool3d`, the argument description of `padding` incorrectly states that zero padding is applied. The remainder of the documentation correctly states that negative infinity padding is applied.
The documentation of `padding` in `nn.MaxPool1d`, `nn.functional.max_pool1d/2d/3d` is correct.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86559
Approved by: https://github.com/albanD
