Use Xoshiro256PlusPlus in examples/rayon-monte-carlo.rs

[dhardy]

Summary

• Add the xoshiro PRNGs in rand::rngs to the local benchmark (redundant with the rngs repo benchmark).
• Remove SmallRng and StdRng from this benchmark (both are thin wrappers around another benchmarked PRNG)
• Adapt the rayon-monte-carlo example to use Xoshiro256PlusPlus instead of ChaCha8Rng

Motivation

Upon consideration of #1781, it occurs to me that there is very little reason to use a thread-local SmallRng instead of a new local one. (In fact, a thread-local one would likely perform worse if we were to use the same per-usage locking as we do in ThreadRng.)

Details

The rayon example now uses multiple seeds instead of a single seed with multiple streams/jumps. I believe this is acceptable. We could instead use jump or jump_n, though (a) we don't have those facilities in the rand::rngs code and (b) I see no reason why this would be preferable. See also the book: parallel RNGs. CC @vigna.