OpenVINO IR branch insertion not working for models converted directly from torch models with OVC

[goodsong81]

Jira ticket: OV fails to build a new model with additional manually constructed branch if model is converted directly from torch

• Working: Torch model -> ONNX export -> OpenVINO import -> IR insertion -> OpenVINO inference
• Not working: Torch model -> OpenVINO conversion -> IR insertion (crash) -> OpenVINO inference

  
  model_xai = ov.Model([*model_ori_outputs, xai_output_node.output(0)], model_ori_params)
  super().__init__(*args)
  RuntimeError: Check 'unregistered_parameters.str().empty()' failed at src/core/src/model.cpp:57:
  Model references undeclared parameters: opset1::Parameter x () -> (f32[?,3,224,224])

Requirements to the input IR: dynamic batch

Thus, currently the OpenVINO XAI "White box" methods which involves IR insertion supports only Torch->ONNX->OpenVINO models.

[goodsong81]

Not reproduced from OpenVINO 2024.2.