Adding this pattern reduces code duplication. There is no need to have a custom implementation for lowering to llvm.cmpxchg. Differential Revision: https://reviews.llvm.org/D78753
