Add JIT fusion pass to fuse quantized add and relu. (#38897)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/38897
Quantized ops support add_relu. This pass enables finding quantized add + relu
pattern and fuse them to add_relu.
Test Plan: buck run caffe2/test:quantization -- test_quantization.TestFusionPasses
Reviewed By: jerryzh168
Differential Revision: D21690909
fbshipit-source-id: 607cf72dde535df15eb7638841543ab2156af464