These entailed constraints only affect performance, and can be removed
without affecting the correctness of the code. They speed up scalar
multiplication of a curve point, used for example in Bitcoinolog.
This allows efficient decryption also of very large files and long
strings.
It is usually highly advisable to stick to common and portable library
predicates such as append/3. However, since append/3 does not yet
recognize this opportunity for improvement, I apply it manually in
this case, so that also very large files can be efficiently decrypted
using the compact internal string representation.
Without this change, decrypting a 1 GB file takes 48 GB of RAM,
whereas with this change, it only takes 2 GB (one for the string, one
for its copy with the appended tag).
Additional authenticated data can now be specified with the new
aad(Chars) option for encryption and decryption. It is authenticated,
but not encrypted.
This avoids the costly (in terms of space requirements!) conversion of
compact lists of characters to lists of integers, making hashing, HKDF,
encryption, decryption, signing and signature verification an order of
magnitude more efficient (primarily in terms of space, also in time).
This makes library(crypto) suitable to process also very large files.
