CPU muon raw to hits optimizations
Reordering by region and quarter is done by linear count sort, not by std::sort.
Multimap is used instead of two nested loops when crossed hits are found.
This MR is as the same as https://gitlab.cern.ch/lhcb-parallelization/Allen/merge_requests/111 but the target branch is master
because plyushchenko-muon-table-reading
branch was merged in https://gitlab.cern.ch/lhcb-parallelization/Allen/merge_requests/90.