normalize before lexicon mapping?

While working on #446 I noticed that normalization doesn't happen before passing the text-to-convert to the lexicon mapping handler. Should we not normalize the text before a lexicon-based mapping too?

See https://github.com/roedoejet/g2p/pull/446#discussion_r2377032956