Add C Extension to optimize building the digest

[DerekStride]

I ran the following benchmark on master and this branch and got the following results.

Benchmark.bmbm do |x|
  x.report('1 000 000') { 1_000_000.times { hexdigest([{foo: 1}]) } }
  x.report('10 000 000') { 10_000_000.times { hexdigest([{foo: 1}]) } }
end

Master

. . . Rehearsal ----------------------------------------
1 000 000    8.350000   0.040000   8.390000 (  8.407654)
10 000 000  80.260000   0.260000  80.520000 ( 80.589933)
------------------------------------ total: 88.910000sec

                 user     system      total        real
1 000 000    7.830000   0.010000   7.840000 (  7.846316)
10 000 000  79.380000   0.240000  79.620000 ( 79.708873)
. . .

This branch with the extension

. . . Rehearsal ----------------------------------------
1 000 000    6.860000   0.010000   6.870000 (  6.882910)
10 000 000  68.250000   0.120000  68.370000 ( 68.382764)
------------------------------------ total: 75.240000sec

                 user     system      total        real
1 000 000    6.840000   0.010000   6.850000 (  6.849427)
10 000 000  68.700000   0.160000  68.860000 ( 68.899472)
. . .