New publications on integer range sampling and shuffling

[vks]

Daniel Lemire published a paper and Melissa O'Neil wrote a blog post about efficient range sampling, including some benchmarks on shuffling performance. This might be of interest for Rand.

[pitdicker]

Just reading the article, we are doing ok already. We use the unbiased multiply method, where the cost of the modulus can be amortized. Or we use the bitmasking method on some integer sizes with sample_single, which may be a bit faster than taking the modulus.

Biased methods are still worth exposing somehow I think.

@vks Thank you for the links!

[dhardy]

It would be interesting to see a straight-up comparison between this and the method @pitdicker implemented, which is similar to Lemire's but uses a different technique to avoid modulo computation. Perhaps @imneme would like to compare?

[vks]

Also see #1154, which implements the O'Neil method for range sampling.