profiler tree test: skip cudaGetDeviceProperties_v2, cudaGetDeviceCount (#107887)
I don't know why these are getting called. But, they only get called on cuda machines, which is breaking tests (https://github.com/pytorch/pytorch/issues/100728).
We can just prune them so that the same result is shown for both CPU and CUDA tests.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107887
Approved by: https://github.com/aaronenyeshi
