Errata around TLS keys in metadata

["Scott Cantor" <cantor.2@osu.edu>]

On various occasions, the question of how to use metadata to determine the
TLS key(s) to accept has come up, and it would be good to nail this down, I
think.

My interpretation for a while has been that a KeyDescriptor with
use="signing" is basically good for either TLS or message security.

I would suggest we codify that, and also clarify that no use attribute
implies that both authentication and key wrapping purposes are acceptable.

-- Scott