Faster 'pruning' of fixations during simulation

• KTfwd::fwdpp_internal::gamete_cleaner has a more efficient implementation now. The worst-case performance of a simulation is the same as in fwdpp 0.4.6, and many reasonable parameter scenarios will seen run times reduced 10-35% (based on the testing that I've done). See fwdpp/internal/gamete_cleaner.hpp for details about how these changes were implemented.
• Various internal improvements to code: const-correctness, better unit tests, better run-time assertions/integration tests.
• A syntax error caught by the Intel compiler was fixed. It never hurt anything on GCC/clang++, but failed to compile using ICC.
• Issue #36 fixed
• Issue #37 fixed