`Sha256Params::new`'s `round` parameter should probably not be a `usize`

[tbu-]

Sha256Params::new

Since the number of rounds has nothing to do with the pointer width of the target CPU, a more concrete type like u32 should probably be used. This is unfortunately a breaking change.

[tarcieri]

I agree a u32 would probably be better. That's what we typically use for round counts in other crates. A breaking change is no problem: we're in the middle of a breaking change cycle.