improve gcd performance

[ahorek]

this patch uses a variant of https://en.wikipedia.org/wiki/Binary_GCD_algorithm

ruby/ruby#1596

master

         Time#subsec      5.377M (±17.9%) i/s -     24.868M in   4.991341s
              Time#-     17.696M (± 9.1%) i/s -     86.842M in   4.982315s
          Time#round      5.162M (±10.1%) i/s -     25.510M in   4.997497s
           Time#to_f     15.985M (± 7.8%) i/s -     78.822M in   4.990285s
           Time#to_r      5.143M (± 8.4%) i/s -     25.479M in   4.993259s
          Rational#+      8.195M (± 4.6%) i/s -     40.802M in   4.995908s
          Rational#-      7.797M (± 5.2%) i/s -     38.806M in   4.999148s
          Rational#*     10.606M (± 4.5%) i/s -     52.840M in   4.994995s

patch

         Time#subsec      6.169M (±18.9%) i/s -     27.650M in   4.983793s
              Time#-     18.941M (± 9.2%) i/s -     92.775M in   4.980521s
          Time#round      5.118M (±10.9%) i/s -     25.194M in   4.993703s
           Time#to_f     17.562M (± 8.2%) i/s -     86.509M in   4.988423s
           Time#to_r      6.094M (± 7.6%) i/s -     30.234M in   4.993605s
          Rational#+      8.519M (± 4.4%) i/s -     42.450M in   4.994747s
          Rational#-      9.363M (± 4.1%) i/s -     46.672M in   4.995156s
          Rational#*     12.275M (± 4.9%) i/s -     61.094M in   4.994644s