almost-passing test for decrypting a Tahoe-LAFS-encoded share

The trick is that we don't use the IV to initialize CTR mode, we only use it
to derive the data encryption key and then use the zero IV to initialize CTR
mode.  (So is it an IV?  Not really I guess.)

The failure is due to incorrect handling of padding somewhere so we get extra
bytes on the recovered ciphertext/plaintext.