[ONNX Importer] QLinearAdd op issue with incorrect return type (!torch.none)

[pravg-amd]

Some of the onnx zoo models have" QLinearAdd" op which results in return type !torch.none on importing the model.

  %88 = torch.operator "onnx.QLinearAdd"(%87, %43, %42, %46, %47, %48, %45, %44) : (!torch.vtensor<[1,4096],ui8>, !torch.vtensor<[],f32>, !torch.vtensor<[],ui8>, !torch.vtensor<[4096],ui8>, !torch.vtensor<[],f32>, !torch.vtensor<[],ui8>, !torch.vtensor<[],f32>, !torch.vtensor<[],ui8>) -> !torch.none

Steps to reproduce:

Build SHARK-TestSuite by following the steps below

https://github.com/nod-ai/SHARK-TestSuite/tree/main/alt_e2eshark

Run the following command

python run.py -va -t bvlcalexnet-12-int8

The model will be available test-run/bvlcalexnet-12-int8/

bvlcalexnet-12-int8
squeezenet1.0-12-int8
caffenet-12-int8
densenet-12-int8
vgg16-12-int8
mobilenetv2-12-int8

[pravg-amd]

This op is not part of https://onnx.ai/onnx/operators/

doc: https://github.com/microsoft/onnxruntime/blob/main/docs/ContribOperators.md#com.microsoft.QLinearAdd

More information regarding this: onnx/onnx#5895

[pravg-amd]

ONNX Runtime ticket to track the support -> microsoft/onnxruntime#24028