Use crypto/rand package for benchmark RNG

[bobertlo]

This is more portable than opening /dev/urandom.

[chmduquesne]

Hi,

Thank you for your contribution. Generally, I like it, however I do not think it makes sense to use crypto/rand in those benchmark tests. The idea is to see how fast those hashes can go through random data, and I am afraid that using a cryptographically secure hash is going to slow down the benchmark quite a bit. This is why I am using /dev/urandom and not /dev/random.

Would you consider switching to math/rand instead? If this is as fast as /dev/urandom I would consider this PR for integration.