[turbofan] Introduce new operators Float32SubPreserveNan and Float64SubPreserveNan for wasm.

src/compiler/mips64/instruction-selector-mips64.cc

Index: src/compiler/mips64/instruction-selector-mips64.cc
diff --git a/src/compiler/mips64/instruction-selector-mips64.cc b/src/compiler/mips64/instruction-selector-mips64.cc
index 0021f1228681fa1f9a94c48a4418f7382cb88827..3516e76501fba1c6793e710bdd3675d1fdb379e6 100644
--- a/src/compiler/mips64/instruction-selector-mips64.cc
+++ b/src/compiler/mips64/instruction-selector-mips64.cc
@@ -1159,6 +1159,9 @@ void InstructionSelector::VisitFloat32Sub(Node* node) {
   VisitRRR(this, kMips64SubS, node);
 }
 
+void InstructionSelector::VisitFloat32SubPreserveNan(Node* node) {
+  VisitRRR(this, kMips64SubS, node);
+}
 
 void InstructionSelector::VisitFloat64Sub(Node* node) {
   Mips64OperandGenerator g(this);
@@ -1178,6 +1181,9 @@ void InstructionSelector::VisitFloat64Sub(Node* node) {
   VisitRRR(this, kMips64SubD, node);
 }
 
+void InstructionSelector::VisitFloat64SubPreserveNan(Node* node) {
+  VisitRRR(this, kMips64SubD, node);
+}
 
 void InstructionSelector::VisitFloat32Mul(Node* node) {
   VisitRRR(this, kMips64MulS, node);