[compiler-rt] [compiler-rt][builtins] Switch to using __builtin_memcpy_inline (PR #69784)

[Saleem Abdulrasool via llvm-commits]

https://github.com/compnerd approved this pull request.

I think that the functional change here is fine and the right thing to do. I worry about compatibility with compilers. Could you please do a follow up (or merge into this change) a change to ensure that the compiler supports this builtin and if not, report a proper error.

https://github.com/llvm/llvm-project/pull/69784