[static runtime] out variant for full_like (#58079)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58079
Support full_like
Test Plan:
`buck test mode/dev caffe2/benchmarks/static_runtime:static_runtime_cpptest -- StaticRuntime.IndividualOps_FullLike`
Test on regenerated local inline_cvr model
```
MKL_NUM_THREADS=1 OMP_NUM_THREADS=1 numactl -m 0 -C 3 ./buck-out/opt/gen/caffe2/caffe2/fb/predictor/ptvsc2_predictor_bench --scripted_model=/data/users/ansha/tmp/adfinder/dec_6x/266377643_shrunk.predictor.disagg.local.regenerated.pt --pt_inputs=/data/users/ansha/tmp/adfinder/dec_6x/local_inputs --pt_enable_static_runtime=1 --pt_cleanup_activations=1 --pt_enable_out_variant=1 --compare_results=1 --iters=5000 --warmup_iters=5000 --num_threads=1 --do_profile=0 --do_benchmark=1 --adsfinder_compatibility=1 --v=1
```
`V0511 10:59:57.187054 1911683 impl.cpp:1229] Switch to out variant for node: %5571 : Tensor = aten::full_like(%blob_for_shape.1, %235, %654, %75, %75, %75, %75)`
Reviewed By: hlu1
Differential Revision: D28361997
fbshipit-source-id: 89c41e37ce23d6008cfe4d80536832ee76d3405e
