[PyTorch] Move selected_mobile_ops.h codegen function to tools (#53786)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53786
To generate `selected_mobile_ops.h` in OSS, move the header file codegen functions to `tools/lite_interpreter/gen_selected_mobile_ops_header.py` file, so OSS can reuse these functions.
ghstack-source-id: 123754437
Test Plan:
```
buck test //xplat/caffe2:supported_mobile_models_test
```
```
buck run //xplat/caffe2:gen_oplist -- --model_file_list_path @/data/users/chenlai/data/pytorch/oplist_folder/file_list_path.macro --allow_include_all_overloads --output_dir /home/chenlai/local/data/pytorch/oplist_folder
```
`file_list_path.macro` content is:
```
chenlai@devvm2090:~/fbsource(45a9b7888)$ cat /data/users/chenlai/data/pytorch/oplist_folder/file_list_path.macro
/data/users/chenlai/fbsource/buck-out/gen/aab7ed39/xplat/caffe2/supported_mobile_models_test_op_list/model_operators.yaml
```
In output folder `/home/chenlai/local/data/pytorch/oplist_folder`, these files are generated:
```
selected_mobile_ops.h selected_operators.yaml SupportedMobileModelsRegistration.cpp
```
the generated files are the same as before.
{P282056731}
{P282055046}
Reviewed By: dhruvbird, iseeyuan
Differential Revision: D26907868
fbshipit-source-id: 9ba786f9c5674a72cad237ae7baadbe4642c51d5