Next round of multireg preliminary changes (dotnet#36155)

This is a zero-diff set of mostly refactoring changes in preparation for supporting multireg locals:
- Move `genRegCopy()` and `genStructReturn()` to codegencommon.cpp, making a new method for `genSIMDSplitReturn` which is target-specific.
- Factor out a new `genUnspillLocal` method from `genUnspillRegIfNeeded()`.
- Similarly factor out `genSpillLocal()`
- Rename `genMultiRegCallStoreToLocal()` and more generally support multireg local stores.
- Fix a bug in the order and shift amount for last-use bits on `GenTreeLclVar`
- Some additional cleanup and preparatory changes

src/coreclr/src/jit/codegen.h

@@ -1113,6 +1113,9 @@ XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
     // Do liveness update for register produced by the current node in codegen after
     // code has been emitted for it.
     void genProduceReg(GenTree* tree);
+    void genSpillLocal(unsigned varNum, var_types type, GenTreeLclVar* lclNode, regNumber regNum);
+    void genUnspillLocal(
+        unsigned varNum, var_types type, GenTreeLclVar* lclNode, regNumber regNum, bool reSpill, bool isLastUse);
     void genUnspillRegIfNeeded(GenTree* tree);
     regNumber genConsumeReg(GenTree* tree);
     void genCopyRegIfNeeded(GenTree* tree, regNumber needReg);
@@ -1275,10 +1278,13 @@ XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
     void genEHFinallyOrFilterRet(BasicBlock* block);
 #endif // !FEATURE_EH_FUNCLETS
 
-    void genMultiRegCallStoreToLocal(GenTree* treeNode);
+    void genMultiRegStoreToLocal(GenTree* treeNode);
 
-    // Deals with codegen for muti-register struct returns.
+    // Codegen for multi-register struct returns.
     bool isStructReturn(GenTree* treeNode);
+#ifdef FEATURE_SIMD
+    void genSIMDSplitReturn(GenTree* src, ReturnTypeDesc* retTypeDesc);
+#endif
     void genStructReturn(GenTree* treeNode);
 
 #if defined(TARGET_X86) || defined(TARGET_ARM)

src/coreclr/src/jit/codegenarm.cpp

@@ -961,7 +961,7 @@ void CodeGen::genCodeForLclVar(GenTreeLclVar* tree)
     // If this is a register candidate that has been spilled, genConsumeReg() will
     // reload it at the point of use.  Otherwise, if it's not in a register, we load it here.
 
-    if (!isRegCandidate && !(tree->gtFlags & GTF_SPILLED))
+    if (!isRegCandidate && !tree->IsMultiReg() && !(tree->gtFlags & GTF_SPILLED))
     {
         const LclVarDsc* varDsc = compiler->lvaGetDesc(tree);
         var_types        type   = varDsc->GetRegisterType(tree);
@@ -1050,7 +1050,7 @@ void CodeGen::genCodeForStoreLclVar(GenTreeLclVar* tree)
     // case is handled separately.
     if (data->gtSkipReloadOrCopy()->IsMultiRegCall())
     {
-        genMultiRegCallStoreToLocal(tree);
+        genMultiRegStoreToLocal(tree);
     }
     else
     {

src/coreclr/src/jit/codegenarm64.cpp

@@ -1839,7 +1839,7 @@ void CodeGen::genCodeForLclVar(GenTreeLclVar* tree)
     // If this is a register candidate that has been spilled, genConsumeReg() will
     // reload it at the point of use.  Otherwise, if it's not in a register, we load it here.
 
-    if (!isRegCandidate && !(tree->gtFlags & GTF_SPILLED))
+    if (!isRegCandidate && !tree->IsMultiReg() && !(tree->gtFlags & GTF_SPILLED))
     {
         // targetType must be a normal scalar type and not a TYP_STRUCT
         assert(targetType != TYP_STRUCT);
@@ -1929,7 +1929,7 @@ void CodeGen::genCodeForStoreLclVar(GenTreeLclVar* tree)
     // case is handled separately.
     if (data->gtSkipReloadOrCopy()->IsMultiRegCall())
     {
-        genMultiRegCallStoreToLocal(tree);
+        genMultiRegStoreToLocal(tree);
     }
     else
     {