Merge pull request #3999 from incubus-ank:fix/cuda_fast_detectAsync_memory_leak

Add missing cudaFree #3999

I found missing memory cudaFree in cv::cuda::FAST_Impl::detectAsync()

I created an issue, but did not receive a response.
#3994

### Pull Request Readiness Checklist

- [x] I agree to contribute to the project under Apache 2 License.
- [x] 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
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable 
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake

Author: incubus-ank

modules/cudafeatures2d/src/fast.cpp

@@ -66,6 +66,7 @@ namespace
     {
     public:
         FAST_Impl(int threshold, bool nonmaxSuppression, int max_npoints);
+        ~FAST_Impl();
 
         virtual void detect(InputArray _image, std::vector<KeyPoint>& keypoints, InputArray _mask);
         virtual void detectAsync(InputArray _image, OutputArray _keypoints, InputArray _mask, Stream& stream);
@@ -95,6 +96,12 @@ namespace
     FAST_Impl::FAST_Impl(int threshold, bool nonmaxSuppression, int max_npoints) :
         threshold_(threshold), nonmaxSuppression_(nonmaxSuppression), max_npoints_(max_npoints)
     {
+        cudaSafeCall( cudaMalloc(&d_counter, sizeof(unsigned int)) );
+    }
+
+    FAST_Impl::~FAST_Impl()
+    {
+        cudaSafeCall( cudaFree(d_counter) );
     }
 
     void FAST_Impl::detect(InputArray _image, std::vector<KeyPoint>& keypoints, InputArray _mask)
@@ -116,8 +123,6 @@ namespace
     {
         using namespace cv::cuda::device::fast;
 
-        cudaSafeCall( cudaMalloc(&d_counter, sizeof(unsigned int)) );
-
         const GpuMat img = _image.getGpuMat();
         const GpuMat mask = _mask.getGpuMat();
 
@@ -165,8 +170,6 @@ namespace
             kpLoc.colRange(0, count).copyTo(locRow, stream);
             keypoints.row(1).setTo(Scalar::all(0), stream);
         }
-
-        cudaSafeCall( cudaFree(d_counter) );
     }
 
     void FAST_Impl::convert(InputArray _gpu_keypoints, std::vector<KeyPoint>& keypoints)