fx quant: remove matching hack for binary qhandler (#57470)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57470
Removes the earlier hack of matching patterns originally matched
to BinaryOpQuantizeHandler to switch to CopyHandler. After this PR,
each pattern can only be matched to one type of QuantizeHandler or
to nothing.
Test Plan:
```
python test/test_quantization.py TestQuantizeFx
python test/test_quantization.py TestQuantizeFxOps
```
Imported from OSS
Reviewed By: jerryzh168
Differential Revision: D28152909
fbshipit-source-id: afc285e770bd7eb0518c90e3ee4874c421e78bbc
