[cudawarping] Add align_corners option to cv::cuda::resize

[zhangnju]

Summary

cv::cuda::resize with INTER_LINEAR currently uses the scale mapping src = dst * srcSize/dstSize and rounds the interpolated value to nearest. This PR adds an optional align_corners flag (INTER_LINEAR only) that instead maps src = dst * (srcSize-1)/(dstSize-1) and truncates toward zero.

This matches PyTorch's interpolate(mode="bilinear", align_corners=True) and the reference integer bilinear used by several ML image preprocessors (e.g. Qwen-VL), which is useful when a GPU preprocessing path must be numerically consistent with a CPU/reference pipeline.

Details

• New bool align_corners = false parameter on cv::cuda::resize (default keeps existing behaviour; public InterpolationFlags enum unchanged).
• Implemented via a new truncating-linear kernel resize_linear_trunc, selected through an internal interpolation sentinel; the default INTER_LINEAR path is untouched.
• Verified element-wise against the CPU reference bilinear: max abs diff = 1/255 (a single quantization step), zero elements off by more than one quantum.

Pull Request Readiness Checklist

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
• The PR is proposed to the proper branch
• There is a reference to the original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable