Random 4K needs to use a good random number generator

[gflofst]

Here is a pure C implementation of a Mersenne Twister:
https://github.com/ESultanik/mtwister?tab=readme-ov-file

We should use this rather than rand and srand in the code.

[adilger]

@gflofst do you have any indication that the rand() function is sufficiently bad that it cannot be used for the IOR random read phase?

According to the Linux rand(3) man page, it is using the same internal implementation as random(3), which reports a periodicity of 16 x ((2^31)−1) or about 34B numbers, which is totally sufficient for a single client's reads, which are currently capped at 1M. Honestly, I don't think there even is a non-Linux system in the Top-500 so I don't think poor rand() implementations on ancient SunOS or AIX versions matters.