A non-empty payload of length a multiple of the chunk size can be
encrypted in two ways: with the last chunk full, or with an extra empty
last chunk. This is mostly an oversight in the original spec.
Both age and rage generate full last chunks, so we should be still in
time to pick one of the two, and avoid the underspecification. It's not
the one I would have picked originally, maybe, because disallowing full
last chunks would have avoided the trial decryption, but oh well.
The caller can take care of the armor. For consistency move the
responsibility to close the armor to the caller, and make the stream
Writer not propagate Close.
This also will also allow us to spin the armor implementation out into
its won package that imports format, without getting an import loop from
format.Parse magically invoking armor decoding.
Less magic in the API, more magic in the CLI.
