Remove `AffineQuantizer.h` from `qnms_kernel.cpp` (#6141)

As @vfdev-5 proposed in https://github.com/pytorch/vision/issues/6138#issuecomment-1150398081

Co-authored-by: vfdev <vfdev.5@gmail.com>

torchvision/csrc/ops/quantized/cpu/qnms_kernel.cpp

 #include <ATen/ATen.h>
-#include <ATen/native/quantized/AffineQuantizer.h>
 #include <torch/library.h>
 
 namespace vision {