Fix two issues with grandfathered tags

[littledan]

cc @anba @gsathya

[anba]

This doesn't like right, because it means new Intl.Locale("zh-min", {language: "zh-min-nan", region: "CN"}).toString() returns "nan-CN", whereas new Intl.Locale("zh", {language: "zh-min-nan", region: "CN"}) will throw a RangeError.

I think we should keep step 4.b and throw a RangeError if the "language" option is a grandfathered tag.

Also kind of related, #50 made ApplyOptionsToTag more complicated, because now it's necessary to detect and throw a RangeError for things like new Intl.Locale("en-gaulish", {language: "cel", region: "FR"}) ("en" is replaced with "cel", which changes the tag to "cel-gaulish", so a grandfathered tag. And then the current spec requires to throw a RangeError when attempting to add the region "FR" to this grandfathered tag.)

[littledan]

We agreed on this in https://github.com/tc39/ecma402/blob/master/meetings/notes-2018-08-16.md but I need to take a closer look at the issue @anba raised above.

[littledan]

I think we should keep step 4.b and throw a RangeError if the "language" option is a grandfathered tag.

If I understood correctly, @srl295 has been arguing for nicer handling in edge cases like these. I wonder what he thinks about @anba's proposal.

Also kind of related, #50 made ApplyOptionsToTag more complicated, because now it's necessary to detect and throw a RangeError for things like new Intl.Locale("en-gaulish", {language: "cel", region: "FR"}) ("en" is replaced with "cel", which changes the tag to "cel-gaulish", so a grandfathered tag. And then the current spec requires to throw a RangeError when attempting to add the region "FR" to this grandfathered tag.)

Do you have a recommendation for what we should do in a case like this?

I'm really not happy about all the complexity that we're taking in due to grandfathered tags; it seems like we're doing a lot of work for edge cases where there's been a specific decision not to focus on an older representation. Is there any way we can avoid this?

[littledan]

Would be obsoleted by #66

[littledan]

OK, I think this PR is subsumed by the later work about Unicode BCP 47 Locale Identifiers.