[llvm] [NewPM] Add port for aarch64-simd-scalar (PR #185256)

[Aiden Grossman via llvm-commits]

================
@@ -62,9 +62,9 @@ STATISTIC(NumCopiesInserted, "Number of cross-class copies inserted");
 #define AARCH64_ADVSIMD_NAME "AdvSIMD Scalar Operation Optimization"
 
 namespace {
-class AArch64AdvSIMDScalar : public MachineFunctionPass {
-  MachineRegisterInfo *MRI;
-  const TargetInstrInfo *TII;
+class AArch64AdvSIMDScalarImpl {
----------------
boomanaiden154 wrote:

And if we want to get rid of the `Impl` suffix, we should do it across X86/AMDGPU/ARM/AArch64. There are tons of ways that different targets differ for no good reason, and I'd rather not add more.

https://github.com/llvm/llvm-project/pull/185256