Align DateError with DateFromFieldsError

[robertbastian]

#7010

Adds DateError::MonthCodeNotInCalendar, DateError::MonthCodeNotInYear, docs, and flattens DateFromFieldsError::Range

[sffc]

I'm not aligned on this, as discussed in the issue.

[Manishearth]

As it stands this is a breaking change, yes?

[robertbastian]

no

[Manishearth]

Ah, I see, UnknownMonthCode is still around. There is a little bit of concern around splitting the error variants but I don't really think matching on these variants is that common.

I like this. I'm pretty ambivalent on the "nested RangeError or struct variant" thing, our design here has some mistakes but overall this is an extremely minor aspect of the API.

[sffc]

I'm against adding new variants to DateError if it causes existing call sites with non-exhaustive matching on error variants to start falling into the catchall variant.

And my position on the nested RangeError hasn't changed since last discussed in #7010.