Polymorphic transcript function does not allow for different prefixes when absorbing different type of elements

[iquerejeta]

In the previous implementation the transcript interface had a common_scalar and common_point interface. This allowed to have different prefixes when adding each type of arguments to the transcript (see here) as follows:

/// Prefix to a prover's message containing a curve point
const BLAKE2B_PREFIX_POINT: u8 = 1;

/// Prefix to a prover's message containing a scalar
const BLAKE2B_PREFIX_SCALAR: u8 = 2;

then, the hasher was initialised with

self.state.update(&[BLAKE2B_PREFIX_POINT]);

in case of a point and with

self.state.update(&[BLAKE2B_PREFIX_SCALAR]);

in case of a scalar.

Now we use the same prefix for every type of input. I believe that is fine, as the goal of these prefixes is to have a separator for every new input. However, I did want to check that I'm not missing out anything, and that it is indeed crucial that those prefixes are distinct.