[test] Use numpy's tolerance for float16#491

brandon-b-miller merged 1 commit intoNVIDIA:mainNVIDIA/numba-cuda:mainfrom

ashermancinelli:ajm/allclose-utilityashermancinelli/numba-cuda:ajm/allclose-utilityCopy head branch name to clipboard