Spun off from https://github.com/tc39/ecma402/issues/273
CANONICALIZE_MAP in uloc.cpp has quite a lot of entries for legacy/bogus/non-registered locale ids that are mapped to more regular locale ids.
e.g. es_ES_PREEURO to es_ES@currency=esp, zh_Hans_PINYIN to zh_Hans@collation=pinyin, etc.
When implementing Ecma 402's locale related APIs with ICU, those entries need to be filtered out. Having a strict compliance mode for BCP 47 + RFC 6067 (+ IANA language subtag registry) would be very helpful.