…Date

Similar to the previous commit, it's possible to get here without
_fields_.[[Month]] if the property bag contained `monthCode` and `day`.

h/t fabon-f

This needs to parallel the non-chinese/dangi case.

In this step we only need the month code, so there's no need to use the
given day (which may be 30 even in a 29-day month, in which case the
normal completion assertion would fail.)

Make the same change in the non-chinese/dangi case even though it's not
strictly necessary.

h/t fabon-f

This makes the chinese/dangi and non-chinese/dangi clauses mirror each
other a bit better.

Makes the behaviour of the chinese/dangi clause correspond to the
non-chinese/dangi clause. This fixes an unintended regression from tc39#118.

h/t fabon-f

After having fixed the assertion failures etc. in the previous commits,
these two clauses look a lot more alike. It's possible to combine them,
only special-casing two things for chinese/dangi: the maximum number 30
of days in month, and the extra lookup in the table to handle rare
month codes.

Closes: tc39#123

I find the "If X is not UNSET, Y is not UNSET" hard to reason about. We
already have an if-clause that covers these two code paths, so we can
move the assertions into the beginning of the if and else branches.