cryptodev: Support ciphers with blocksize != ivsize

[cem]

Summary

For example, Chacha20 expects a 16 byte IV+nonce but is a stream cipher with
an effective input block size of a single byte.

This change is the bare minimum; some inefficiencies remain in
cryptosoft.c. I plan to address those once there is a cipher to test
them against.

[cem]

The transform's ivsize is how large an IV is — not its blocksize. Match code with intent.

Summary of existing enc algs:

• Skipjack, AES-ICM, Blowfish, CAST-128, Camellia, DES3, Rijndael128, DES: Identical IV and block size (no functional change).

• AES-GCM: block len of 1, IV len of 12.

• AES-XTS: block len of 16, IV len of 8.

• NULL: block 4, IV 0

For all of these it seems like the supplied ivsize is correct, in the cases where it makes a difference.

Also, for Chacha20 (not yet added to OCF): block len 1, IV len 16 (IV + Nonce anyway).

Does this make sense?

[cem]

r331579