0-dim batch size input for interpolate. (#32400)
Summary:
This PR adds support for 0-dim batch size input for `torch.nn.functional.interpolate` for various modes of interpolation.
Fixes part of gh-12013
CC: rgommers ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/32400
Differential Revision: D19557090
Pulled By: ezyang
fbshipit-source-id: 6822f148bb47bfbcacb5e03798bf2744f24a2a32
