Lower smulo codegen opcode

[vladimirradosavljevic]

For the following test:

target datalayout = "E-p:256:256-i256:256:256-S32-a:256:256"
target triple = "eravm"

declare { i256, i1 } @llvm.smul.with.overflow.i256(i256, i256)

define i256 @test(i256 %a, i256 %b, i256 %x, i256 %y) {
entry:
  %smul = call {i256, i1} @llvm.smul.with.overflow.i256(i256 %x, i256 %y)
  %overflow = extractvalue {i256, i1} %smul, 1
  %select = select i1 %overflow, i256 %a, i256 %b
  ret i256 %select
}

we are currently hitting the assert (Assertion failed: (LC != RTLIB::UNKNOWN_LIBCALL && "Cannot expand this operation!"), function expandMULO), since we don't support lowering of smulo codegen opcode.
Possible solution is here.