[RISCV] Add TU variants to VPseudoBinaryM. NFC

[luke]

Summary

The aim is to use these in https://reviews.llvm.org/D152963 so that we
can preserve the passthru operand whilst also emitting the unmasked
variant. Like the mask variant, they are marked with forceTailAgnostic.

[craig.topper]

Is BinaryM only used by compares? The name is not super helpful.

[luke]

In D153011#4425334, @craig.topper wrote:

Is BinaryM only used by compares? The name is not super helpful.

Yeah, I think so. Should I rename it to VPsuedoVCMPM to match VPseudoVCMPM_VV_VF et al?

[luke]

After the pseudo refactoring this may no longer be relevant

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

 multiclass VPseudoBinaryM<VReg RetClass,
                           VReg Op1Class,
                           DAGOperand Op2Class,
                           LMULInfo MInfo,
                           string Constraint = ""> {
   let VLMul = MInfo.value in {
     def "_" # MInfo.MX : VPseudoBinaryNoMask<RetClass, Op1Class, Op2Class,
                                              Constraint>;
-    let ForceTailAgnostic = true in
+    let ForceTailAgnostic = true in {
+      def "_" # MInfo.MX # "_TU" : VPseudoBinaryNoMaskTU<RetClass, Op1Class,
+                                                         Op2Class, Constraint>;
       def "_" # MInfo.MX # "_MASK" : VPseudoBinaryMOutMask<RetClass, Op1Class,
                                                            Op2Class, Constraint>,
-                                   RISCVMaskedPseudo</*MaskOpIdx*/ 3, /*HasTU*/ false>;
+                                     RISCVMaskedPseudo</*MaskOpIdx*/ 3>;
+    }
   }
 }
 
 multiclass VPseudoBinaryEmul<VReg RetClass,
                              VReg Op1Class,
                              DAGOperand Op2Class,
                              LMULInfo lmul,
                              LMULInfo emul,