[ONNX] Support ListConstruct in quantized_args (#92009)
Fixes #91303
quantized_args didn't support ListConstruct leading to an error when user uses quantized op with list inputs, ex: aten::cat. After this PR, converter can successfully export the issued model and pass ONNX checker. However, ORT doesn't seem to support it with the very same error as https://github.com/microsoft/onnxruntime/issues/12131.
Update:
I find test_quantized_cat_when_concatinating_the_same_tensor is even similar to the new case we have in here. The only difference is whether the inputs are already quantized. ONNX graphs both seem to be valid.
[test_quantized_cat_when_concatinating_the_same_tensor.zip](https://github.com/pytorch/pytorch/files/10396798/test_quantized_cat_when_concatinating_the_same_tensor.zip)
[test_quantized_list_of_inputs_with_cat.zip](https://github.com/pytorch/pytorch/files/10396799/test_quantized_list_of_inputs_with_cat.zip)
issue raised https://github.com/microsoft/onnxruntime/issues/14245
Pull Request resolved: https://github.com/pytorch/pytorch/pull/92009
Approved by: https://github.com/BowenBao
