Adds the benchmarks from https://github.com/spenczar/tdigest
(Specifically, incorporates benchmarks from https://github.com/spenczar/tdigest/blob/master/benchmarks_test.go )
For what it's worth, the benchmarks aren't a great representation of real-world performance in my hands. Although the benchmarks seemingly favor tdigest by a wide margin, in actual use the streadway/quantile package had much better performance for the exact same task. The underlying task takes 440 seconds. The tdigest package added 4,000 seconds of overhead, while the streadway/quantile package added 2,800 seconds of overhead. The quantile compute time was dominated by sorting, while the tdigest compute time was dominated by waiting for a mutex.