Fix attention fusion for UNet onnx model export when using LoRA weigh…

…ts (microsoft#17249)

### Description
Tested with stable diffusion unet models exported by both pytorch 2.1.0
(nightly) and pytorch 1.13.1, with and without LoRA weights.



### Motivation and Context
LoRA weights modifiy the unet model by adding matmul and scale
operations to every q/k/v/out tensors, which breaks the current MHA
pattern recognition.