BEEFY: Simplify hashing for pallet-beefy-mmr #12387

Right now pallet-beefy-mmr Declares a new Hasher trait and a new Keccak256 struct which implements it. I was wondering if this is needed.

I was thinking of 2 possible simplifications:

1. Instead of defining a new struct we could reuse the Keccak256 struct that's already present in sp_runtime, doing something like:

#[cfg(feature = "keccak")]
mod keccak256 {
	pub use sp_runtime::traits::Keccak256;

	/// Keccak256 hasher implementation.
	impl super::Hasher for Keccak256 {
		fn hash(data: &[u8]) -> super::Hash {
			Self::hash(data).into()
		}
	}
}

Or we could do even something more generic:

	impl<T> super::Hasher for T
		where T: sp_runtime::traits::Hash,
		<T as sp_runtime::traits::Hash>::Output: Into<super::Hash> {
		fn hash(data: &[u8]) -> super::Hash {
			Self::hash(data).into()
		}
	}

2. Could we go one step further and remove the Hasher trait completely and use sp_runtime::traits::Hash instead ? And maybe in the end we could also remove this.

@acatangiu WDYT ? If this issue makes sense I would be happy to implement it.