This is a feature request I would be willing to implement if it is deemed appropriate.
I often find myself wanting to generate large pseudo-randomly filled files for testing
and benchmarks. The io.Reader interface in crypto/rand is nice, but it is
non-deterministic by design. Having an io.Reader interface on math/rand.Rand, paired
with a consistent seed, would allow consistent random number generation from run-to-run
of tests and benchmarks.
Would this be an appropriate addition to the standard library?