core(ocl): fix deadlock in UMatDataAutoLock

UMatData locks are not mapped on real locks (they are mapped to some "pre-initialized" pool).

Concurrent execution of these statements may lead to deadlock:
- a.copyTo(b) from thread 1
- c.copyTo(d) from thread 2
where:
- 'a' and 'd' are mapped to single lock "A".
- 'b' and 'c' are mapped to single lock "B".

Workaround is to process locks with strict order.

modules/core/src/ocl.cpp

@@ -123,6 +123,8 @@
 #include "opencv2/core/opencl/opencl_svm.hpp"
 #endif
 
+#include "umatrix.hpp"
+
 namespace cv { namespace ocl {
 
 #define IMPLEMENT_REFCOUNTABLE() \
@@ -5424,8 +5426,7 @@ public:
                                             srcrawofs, new_srcofs, new_srcstep,
                                             dstrawofs, new_dstofs, new_dststep);
 
-        UMatDataAutoLock src_autolock(src);
-        UMatDataAutoLock dst_autolock(dst);
+        UMatDataAutoLock src_autolock(src, dst);
 
         if( !src->handle || (src->data && src->hostCopyObsolete() < src->deviceCopyObsolete()) )
         {