* Add DeformableConv2D
* add unittest and docs
* pnnx torchvision deformconv2d conversion

Co-authored-by: miemie2013 <miemie2013@users.noreply.github.com>
Co-authored-by: nihui <shuizhuyuanluo@126.com>

* implement 4d memorydata

* fix ncnnoptimize memorydata 4d

* clear bom marker, avoid vector data function

* convert inplace relu6, match more hardswish pattern
* upgrade requests

* insert reshape for nn.Linear with 4d/5d inputs

* batch index aware reshape, add test

src/stb_image.h
src/layer/detectionoutput.cpp
tools/quantize/imreadwrite.cpp

* fix crash in ncnn2table for missing weight data in lightmode

* Update modelwriter.h

* disable mkl avx/avx2 optimization for potential memory issue in dft

* chain dict output

* recognize sub_

* feat(src/layer): add vision_transformer benchmark and relative layer

* refactor(testutil.h): add para for RandomMat

* bitwise and compare op

* masked_fill and gather test

* normalize einsum equation

* replace dot pair with alphabet word

* shorten permute equation

* convert torch.bmm

* convert torch.min torch.max