move win cuda tests from pr to trunk
Fixes #ISSUE_NUMBER
helps w/ #76838
as in title
they take a long time theres a lot of queuing for the windows.8xlarge.nvidia.gpu machines, hopefully this will bring the queue time down + decrease tts
avg tts for the past week for `pull / win-vs2019-cuda11.3-py3 / test` is 4.7 and 4.4 hours
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76909
Approved by: https://github.com/seemethere, https://github.com/suo, https://github.com/janeyx99
