Improvement: Use Twofish approach for test vectors

[Legrandin]

It would be nice if ciphers (or even other complex primitives, like hashes) had test vectors generated as described by Niels Ferguson here:

http://blogs.msdn.com/b/si_team/archive/2006/05/19/aes-test-vectors.aspx

===
Notation:

            b # bytes in a plaintext or ciphertext block

            k # bytes in a key

            E(K,P) Block cipher encryption function, K = key, P = plaintext

            D(K,P) Block cipher decryption function, K = key, C = ciphertext

Test algorithm:

    S = a string of k+b zero bytes.

    repeat 1000 times:

            n = length(S)

            K = S[n-k..n-1] the last k bytes of S

            P = S[n-k-b..n-k-1] the b bytes just before K

            append E(K,E(K,P)) to S

The last b bytes of S are the test vector value.
===