[quant][graphmode][refactor] getGeneralTensorInputs(Node*) -> getPassThroughInputs(Value*) (#35558)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/35558
This is to have a more fine grained support for general ops,
e.g. for sort, the first output will have pass through inputs and the second op does not need to be quantized
so we'll have a check for that
Test Plan:
.
Imported from OSS
Differential Revision: D20752128
fbshipit-source-id: 825c4c393910a88ecb12e24e9a2f3b05c5d5a7ab
