Make fb::sigrid_hash_compute_multipler_shift return a std::tuple<int64_t, int64_t> (#67123)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67123
Makes `sigrid_hash_compute_multipler_shift` return a tuple instead of a tensor and modifies functions that depends on it.
Test Plan:
```
buck test //caffe2/benchmarks/static_runtime/fb:test_fb_operators
```
Benchmarks:
`local`:
```
I1022 13:56:34.529495 2866038 PyTorchPredictorBenchLib.cpp:266] Mean milliseconds per iter: 5.67114, standard deviation: 0.336918
I1022 15:29:45.248790 3292725 PyTorchPredictorBenchLib.cpp:266] Mean milliseconds per iter: 5.66678, standard deviation: 0.403032
```
`local_ro`:
```
I1022 13:59:24.262511 2882599 PyTorchPredictorBenchLib.cpp:266] Mean milliseconds per iter: 1.56012, standard deviation: 0.0537101
I1022 15:34:53.941890 3328358 PyTorchPredictorBenchLib.cpp:266] Mean milliseconds per iter: 1.5525, standard deviation: 0.0280267
```
FB: local - P463676888, local_ro - P463676984, master local - P463686094, master local_ro - P463686470
Reviewed By: mikeiovine
Differential Revision: D31867186
fbshipit-source-id: 0f640487b74d1cd0d5f714f2258e056a2f0c2c07