MobileNet-w4a4 accuracy in README looks incorrect

[maltanar]

The top-1 accuracy in MobileNet-w4a4 reported in the README here is 84.22%:

https://github.com/fastmachinelearning/QONNX_model_zoo/tree/main/models/ImageNet/Brevitas_FINN_mobilenet

...while in the Brevitas repo it is 71.14%:

https://github.com/Xilinx/brevitas/tree/master/src/brevitas_examples/imagenet_classification

I believe the latter one (Brevitas) is correct.