[ONNX] Delete redundant dynamic dim check for result types (#3942)

The dynamic has been supported by the code, the check is useless and will block the 
dynamic examples. This will fix nod-ai/SHARK-ModelDev#893

lib/Conversion/TorchOnnxToTorch/DefaultDomainQtoZ.cpp

@@ -4487,11 +4487,6 @@ void mlir::torch::onnx_c::populateDefaultDomainQtoZ(
         SmallVector<Type> scanOutTypes;
         for (unsigned i = numInits; i < resultTypes.size(); i++) {
           auto scanOutTy = cast<Torch::ValueTensorType>(resultTypes[i]);
-          // TODO: Handle dynamic result types.
-          if (!scanOutTy.hasSizes() || !scanOutTy.areAllSizesKnown()) {
-            return rewriter.notifyMatchFailure(
-                binder.op, "Expects result type to be static");
-          }
           Value sizeList =
               createConstantIntList(binder, rewriter, scanOutTy.getSizes());
           initVals.push_back(Torch::createInitTensor(rewriter, loc, scanOutTy,