Merge pull request #16097 from YashasSamaga:cuda4dnn-optimize-resize-bilinear

cuda4dnn(resize): process multiple channels each iteration

* resize bilinear: process multiple chans. per iter.

* remove unused headers

* correct dispatch logic

* resize_nn: process multiple chans. per iter.

modules/dnn/src/cuda/fill.cu

@@ -32,7 +32,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace kernels {
         }
     }
 
-    template <class T, std::size_t N>
+    template <class T, std::size_t N> static
     void launch_vectorized_fill(const Stream& stream, Span<T> output, T value) {
         CV_Assert(is_fully_aligned<T>(output, N));