Migrate bce loss from CUDA_tensor_apply3 to TensorIterator (#34023)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/34023
Test Plan: Imported from OSS
Differential Revision: D20196084
Pulled By: VitalyFedyunin
fbshipit-source-id: bd000f09139cb848562e5310f10067db85e1b935