ArithToLLVM
arith.select
Add 1:N support for the `arith.select` lowering. Only cases where the entire true/false value is selected are supported.