Add a way to specify a control report

[davy]

For some reports, a control report would be nice to be able to separate out expensive, shared behavior.

For example, the 'dup' report in this benchmark:

report = Benchmark.bigo do |x|

  x.generator {|size|
    (0...size).each_with_object({}) do |i, h|
      h[i] = SecureRandom.hex
    end
  }

  # dup acts as a control for this test
  # because merge! modifies the existing hash,
  # the hash needs to be duped in order to test
  # appropriately
  #
  # this means that to control for this work in the
  # other reports, the hash must also be duped, even
  # for merge where the original hash is not modified
  x.report("dup") { |items, size|
    items.dup
  }
  x.report("merge") { |items, size|
    items.dup.merge(rand(size) => SecureRandom.hex)
  }
  x.report("merge!") { |items, size|
    items.dup.merge!(rand(size) => SecureRandom.hex)
  }
  x.report("set") { |items, size|
    items.dup[rand(size)] = SecureRandom.hex
  }
end

In this case, if the expense of the 'dup' report were then subtracted from the results of the other reports, a cleaner view of the differences in behavior would be possible.