The `TooLong` validation message mentioned characters, but the `len`
function actually measures bytes, no characters. This switches it over
to use bytes.
Characters are mostly an illusory concept anyway -- a vain attempt to
shield our minds against the lovecraftian nightmares that comes from
attemping to truly comprehend that eldritch treatise known as the
Unicode standard. Which is to say: measuring things in characters is
hard and mostly ambiguous, and probably not what we meant to do.