getoffsetnanosecondsfor tests might be inconsistently passing/failing

[12wrigja]

While debugging CI failures in #210 , it seems that only some of the test262 cases for toLocaleString/timezone-getoffsetnanosecondsfor-not-callable.js fail, and some pass:

Passing:
PlainTime
PlainDate
PlainDateTime

Failing:
PlainYearMonth
PlainMonthDay
DateTimeFormat/prototype/format
DateTimeFormat/prototype/formatToParts
DateTimeFormat/prototype/formatRange
DateTimeFormat/prototype/formatRangeToParts

I haven't looked at all the testcases but given they are likely all very similarly structured I'm wondering this is indicative of a bug somewhere?

Failing that, it would be nice to improve the documentation to explain why those tests fail on Node<16 more clearly.

[justingrant]

Is the difference that the passing ones are calling toLocaleString with no arguments, while the failing ones are calling with these arguments?

instance.toLocaleString(undefined, { calendar: "iso8601" })

It may be that Node 14 didn't support the calendar option, so it's failing with a RangeError before the uncallable thing is called?

So it's not like these tests run without exceptions in any case. It's only that the ones with a calendar option fail in an unexpected way (RangeError) on Node 14, while the ones without a calendar option fail in an expected way (TypeError).

Anyway, I think it's OK to just put the failing tests into the expected-failures-before-node16.txt and leave the passing tests (the ones that fail with the expected TypeError) out of that file.