Fix invalid threading of nodes in rationalization (#64012)

The code in question assumes that the ASG will be
reversed and thus threads "simdTree" before "location"
in the linear order. That dependency, while valid,
because "gtSetEvalOrder" will always reverse ASGs
with locals on the LHS, is unnecessary and incorrect
from the IR validity point of view.

Fix this by using "InsertAfter" instead of manual
node threading.

src/coreclr/jit/rationalize.cpp

@@ -416,13 +416,10 @@ void Rationalizer::RewriteAssignment(LIR::Use& use)
                 }
                 GenTreeSIMD* simdTree =
                     comp->gtNewSIMDNode(simdType, initVal, SIMDIntrinsicInit, simdBaseJitType, genTypeSize(simdType));
-                assignment->AsOp()->gtOp2 = simdTree;
-                value                     = simdTree;
-                initVal->gtNext           = simdTree;
-                simdTree->gtPrev          = initVal;
+                assignment->gtOp2 = simdTree;
+                value             = simdTree;
 
-                simdTree->gtNext = location;
-                location->gtPrev = simdTree;
+                BlockRange().InsertAfter(initVal, simdTree);
             }
         }
 #endif // FEATURE_SIMD