[MLIR][TORCH] Fix typeConverted automatically convert the ui8 to i8

  When converting vtenor to buildin tensor, the converter wrong convert the ui8 to i8.
  This will lead to other type cast issue in later usage of i8. Here is example:
  nod-ai/SHARK-Turbine#110

lib/Dialect/Torch/IR/TorchTypes.cpp

@@ -405,8 +405,7 @@ static Type convertDtypeToBuiltinElementType(MLIRContext *context, Type dtype) {
   if (auto floatType = dtype.dyn_cast<mlir::FloatType>()) {
     return dtype;
   } else if (auto integerType = dtype.dyn_cast<IntegerType>()) {
-    return IntegerType::get(context, integerType.getWidth(),
-                            IntegerType::Signless);
+    return dtype;
   } else if (dtype.isa<mlir::ComplexType>()){
     return dtype;
   }