Provide deterministic API

[koraa]

It would be useful if pqclean exposed deterministic versions of KEMs and Signature Algorithms taking explicit coins/secret seed parameters. These would come in handy because some protocols mandate their use 0 and they would provide protocol authors with the ability to easily arrive at and check test vectors for their protocols.

[thomwiggers]

I agree completely that this would be extremely useful. It would also have avoided all the messiness that we've had to build around the different implementations that we have of randombytes().

However, the reality is that the NIST API hasn't provided this, so offering this isn't trivial and would requiring significant effort to modify every single scheme.