small cleanup in unicode_norm.c

We've had get_canonical_class() for a while as a backend-only function. There is some ad-hoc code elsewhere that implements the same logic in a couple places, so it makes sense for all sites to use this function instead, as in the attached.
--