with __build_popcountl doesn't return correct number of bits. the ll version should work, but for simplicity,
locking it down to long for now.  TODO add a unit test for the Hamming distance.

--no commit message

updated main CMakeLists.txt gpu module section, now user can manage binary and intermediate code versions of the gpu module image
added more functions to check version of gpu code in runtime

fixing bug related to using the hamming distance on descriptors whose length is not divisible by sizeof(size_t).