minor build system changes (now cuda code in opencv_core is compiled using CUDA_ARCH* cmake variables)

Fixed build for x64 with sse2 disabled; several improvements in performance testing framework; enabled SANITY_CHECK() for performance tests

Added possibility to configure TBB behavior in Perf tests --- now we can set the number of TBB threads in the perf tests either as console parameter perf_tbb_nthreads or 
in a test itself using the method declare.tbb_threads(n).