Improve the precision of abs() and sign() for large values (#99550)
@ev-br found in
https://github.com/Quansight-Labs/numpy_pytorch_interop/pull/117#issuecomment-1514959633
that the precision of `abs()` for large values in the vectorised case is less-than-good.
This PR fixes this issue. While doing that, we are able to comment out a
few tests on extremal values.
Fixes https://github.com/pytorch/pytorch/issues/53958 https://github.com/pytorch/pytorch/issues/48486
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99550
Approved by: https://github.com/ngimel, https://github.com/peterbell10
