used new CV_PERF_TEST_MAIN_WITH_IMPLS mechanism

Turns out, you can't use preprocessor directives inside macro arguments.
Who'd have thought?

Now it allows choosing between arbitrary implementation variants.

now it executes only on one device
added posibility to specify device on which tests will be executed

removed void integral(const GpuMat& src, GpuMat& sum, GpuMat& sqsum, Stream& stream) - it fails with NPP_NOT_IMPLEMENTED error
updated docs, accuracy and performance tests

Perf tests: method ::perf::TestBase::getDataPath made public static; added base CUDA detection into run.py script