Implement a different PRNG for deriving challenge for NI proofs

[survived]

Problem with current approach of deriving a challenge is that it truncates entropy to 256 bits. Depending on choice of security level, implementation might need a higher security level. Suggested approach to address that is to have our own implementation of PRNG (RngCore) that calculates i-th chunk of 32 random bytes as H(i || ..args).

[survived]

I see two ways to implement this PRNG:

1. Accept seed: Fn(Digest) -> Output<Digest> \ i-th chunk of randomness would be

   seed(Digest::new().chain(i))

2. Accept seed: &[&[u8]] \ i-th chunk of randomness would be

   Digest::new()
      .chain(i)
      .chain(seed[0])
      /* ... */
      .chain(seed[n-1])
      .finalize()

[survived]

Let me know if you have other ideas @d86leader