x64: Add native lowering for scalar fma (#4539)

Use `vfmadd213{ss,sd}` for these lowerings.

cranelift/codegen/src/isa/x64/inst/args.rs

@@ -1383,6 +1383,8 @@ impl fmt::Display for SseOpcode {
 
 #[derive(Clone, PartialEq)]
 pub enum AvxOpcode {
+    Vfmadd213ss,
+    Vfmadd213sd,
     Vfmadd213ps,
     Vfmadd213pd,
 }
@@ -1391,8 +1393,10 @@ impl AvxOpcode {
     /// Which `InstructionSet`s support the opcode?
     pub(crate) fn available_from(&self) -> SmallVec<[InstructionSet; 2]> {
         match self {
-            AvxOpcode::Vfmadd213ps => smallvec![InstructionSet::FMA],
-            AvxOpcode::Vfmadd213pd => smallvec![InstructionSet::FMA],
+            AvxOpcode::Vfmadd213ss
+            | AvxOpcode::Vfmadd213sd
+            | AvxOpcode::Vfmadd213ps
+            | AvxOpcode::Vfmadd213pd => smallvec![InstructionSet::FMA],
         }
     }
 }
@@ -1400,6 +1404,8 @@ impl AvxOpcode {
 impl fmt::Debug for AvxOpcode {
     fn fmt(&self, fmt: &mut fmt::Formatter) -> fmt::Result {
         let name = match self {
+            AvxOpcode::Vfmadd213ss => "vfmadd213ss",
+            AvxOpcode::Vfmadd213sd => "vfmadd213sd",
             AvxOpcode::Vfmadd213ps => "vfmadd213ps",
             AvxOpcode::Vfmadd213pd => "vfmadd213pd",
         };