* fetching google benchmark with cmake * use new thread-safe accumulator in histogram_parallel_filling * update boost, root and gsl benchmarks to use google benchmark