Improve cbor type inclusion detection during codegen (#110)
Previously we werent detecting that the richtext facet features union
needed to have the special union cbor type encoding set on it. This was
because the 'walk' function we use to recursively set this field didnt
cross schema boundaries, and even if we did that shallowly, we could
have already processed the referenced schema and not actually passed the
flag down recursively. So now as a hack i'm just calling the walk-set
function twice. This is obviously bad, and I think we should just
rewrite the codegen code instead of trying to improve it, it was written
before several large lexicon refactors and its super messy as a result.
authored by