Add kernel launch checks in caffe2/aten/src/ATen/native/cuda/ (#49269)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/49269
Added C10_CUDA_KERNEL_LAUNCH_CHECK(); after all kernel launches in caffe2/aten/src/ATen/native/cuda.
Several files in the directory still trigger the check_kernel_launches.py tool. These are false positives as the tool doesn't seem to be parsing MACROS correctly.
Normalization.cuh <- This file is also highlighted by the check_kernel_launches.py tool, but the highlighted regions are device code where exception handling isn't allowed.
Test Plan:
Check that the code still builds with
```
buck build //caffe2/aten:ATen-cu
```
https://pxl.cl/1tLRB
Also ran
```
buck test //caffe2/aten:atest
```
https://pxl.cl/1tLSw
Reviewed By: r-barnes
Differential Revision: D25487597
fbshipit-source-id: 7a6689534f7ff85a5d2262831bf6918f1fe0b745
