Add some more SparseMeta support (#82086)
These I noticed while doing crossref testing on test_sparse with fake
sparse tensors on. The zero_ implementation is self explanatory; the
isinf explicit registration is necessary because the
CompositeExplicitAutograd entry does not actually work for sparse
tensors. This could alternately be alleviated by
https://github.com/pytorch/pytorch/issues/81801
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82086
Approved by: https://github.com/bdhirsh
