Perfect forwarding of arguments to CreateOpNode would be nice

[haved]

The implementation is currently like

template<typename OperatorType, typename... OperatorArguments>
SimpleNode &
CreateOpNode(const std::vector<Output *> & operands, OperatorArguments... operatorArguments)
{
  JLM_ASSERT(!operands.empty());
  return SimpleNode::Create(
      *operands[0]->region(),
      std::make_unique<OperatorType>(std::move(operatorArguments)...),
      operands);
}

but we can avoid potential copying of arguments by using perfect forwarding, like so:

template<typename OperatorType, typename... OperatorArguments>
SimpleNode &
CreateOpNode(const std::vector<Output *> & operands, OperatorArguments&&... operatorArguments)
{
  JLM_ASSERT(!operands.empty());
  return SimpleNode::Create(
      *operands[0]->region(),
      std::make_unique<OperatorType>(std::forward<OperatorArguments>(operatorArguments)...),
      operands);
}

The problem is that this causes some annoying false-positive warnings.
The following snippet from IfConversionTests.cpp, for example:

const auto & c0 = jlm::rvsdg::CreateOpNode<jlm::rvsdg::ctlconstant_op>(
      *gammaNode0->subregion(0),
      jlm::rvsdg::ControlValueRepresentation(0, 2));

gives

tests/jlm/llvm/opt/IfConversionTests.cpp: In function ‘void EmptyGammaWithTwoSubregions()’:
tests/jlm/llvm/opt/IfConversionTests.cpp:180:16: error: possibly dangling reference to a temporary [-Werror=dangling-reference]
  180 |   const auto & c0 = jlm::rvsdg::CreateOpNode<jlm::rvsdg::ctlconstant_op>(
      |                ^~
compilation terminated due to -Wfatal-errors.
cc1plus: all warnings being treated as errors

It is possible to wrap the definitions of CreateOpNode with _Pragma("GCC diagnostic ignored \"-Wdangling-reference\""), but I'm not a huge fan of it. example