[Cryptech Tech] Padding in digest cores

[Rob Austein]

From reading the cases and code for our current digest cores, it looks
like the cores all expect the length count and padding to be handled
in software so that the cores can just work with complete blocks.
Presumably this is deliberate, but, if so, it probably ought to be
mentioned somewhere in the documentation, rather than expecting users
to infer it from hexadecimal test data. :)