change index_select scalar_check to retain dimensionality of input. (#30790)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30790
The index_select documentaiton reads:
"The returned tensor has the same number of dimensions as the original tensor (input)."
But the implementation would return a 0-dimensional tensor iff both the input and index were 0-dimensional.
This change makes it so we retuan a 0-dimensional tensor iff the input is 0-dimensional.
Restacked version of: https://github.com/pytorch/pytorch/pull/30502
Test Plan: Imported from OSS
Differential Revision: D18825717
Pulled By: gchanan
fbshipit-source-id: aeb10c5107e748af3e264fbdc81fff5dd4833cc4
