[PyTorchEdge] Make aten function common to aten and torch_common (#66663)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66663
fb: TensorCompare.cpp is in per-app, a target higher than torch_mobile
Please read this doc to know about [Per-app ATen/native and Template Selective Build](
https://docs.google.com/document/d/1O5--mOAi_gGh2GkE-REo3qJRRQ_Lks69IfgszcB8ThI/edit)
Create a filed called "prim_native_functions.cpp" in ATen, add it to aten_cpu, and cut-paste native::is_nonzero() to prim_native_functions.cpp.
By doing this we move the function to lower layer which is more visible to all targets depending on it.
Instruction count comparison new vs old
https://www.internalfb.com/phabricator/paste/view/P463272302?view=diff
Test Plan:
fb:
```
(base) [pavithran@devvm1803.vll0 /data/users/pavithran/fbsource] buck build //xplat/caffe2:aten_cpu
Building: finished in 0.4 sec (100%) 1/202 jobs, 0/202 updated
Total time: 0.4 sec
More details at https://www.internalfb.com/intern/buck/build/ea35300b-55be-4b9f-bc74-80cdd869c16a
BUILD SUCCEEDED
(base) [pavithran@devvm1803.vll0 /data/users/pavithran/fbsource] buck build //xplat/caffe2:aten_native_cpu
Building: finished in 0.7 sec (100%) 1/1 jobs, 0/1 updated
Total time: 0.8 sec
More details at https://www.internalfb.com/intern/buck/build/ccd97d43-c59d-4f29-9418-485cd24575e2
BUILD SUCCEEDED
```
Reviewed By: iseeyuan
Differential Revision: D31669536
fbshipit-source-id: d35f069f975db6dce0b678c5b5ddd74bd690f599
