[quant] add quantized::batch_norm (#39910)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/39910
We need this for graph mode quantization, since we only have `aten::batch_norm` the dimension
is only known at runtime, we'll need to quantize it to `quantized::batch_norm`
Test Plan: Imported from OSS
Differential Revision: D22012281
fbshipit-source-id: 2973d86a17a02b7bdc36bd1e703e91584d9139d0
