Convert WeChatQRCode test model fixtures from Caffe to ONNX

OpenCV 5.x's wechat_qrcode module (and WeChatQRCode's OpenCvSharp wrapper)
only accepts single-file ONNX models now; the Caffe importer has been
removed entirely. The Caffe-format fixtures checked into this repo
(detect.prototxt/.caffemodel, sr.prototxt/.caffemodel) were therefore stale
and every model-dependent WeChatQRCodeTest case was silently skipped.

Replace them with detect.onnx/sr.onnx, converted from the same published
weights (opencv/opencv_3rdparty:wechat_qrcode_20210119, Apache License 2.0;
see NOTICE.md). All 7 model-dependent tests in WeChatQRCodeTest.cs now run
and pass instead of being skipped.

Refs #2075

Author: shimat

test/OpenCvSharp.Tests/_data/wechat_qrcode/NOTICE.md

@@ -0,0 +1,12 @@
+`detect.onnx` and `sr.onnx` are derived from the WeChatQRCode Caffe models
+published at
+[opencv/opencv_3rdparty:wechat_qrcode_20210119](https://github.com/opencv/opencv_3rdparty/tree/wechat_qrcode_20210119)
+(`detect.prototxt`/`.caffemodel`, `sr.prototxt`/`.caffemodel`), Copyright (C)
+2021 THL A29 Limited, a Tencent company, licensed under the
+[Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0).
+
+These files were converted from Caffe to ONNX using the tool at
+`tools/wechat_qrcode_caffe2onnx/` in this repository, since OpenCV 5.x's
+`wechat_qrcode` module (and OpenCvSharp's `WeChatQRCode` wrapper) only accepts
+ONNX models. No other changes were made to the network weights or
+architecture.