CharacterDirectionOfLocale: Define if locales are maximized before lookup

[anba]

The V8 implementation of CharacterDirectionOfLocale calls uloc_isRightToLeft and uloc_isRightToLeft adds likely subtags when no explicit script subtag is present. Notably likely subtags aren't added even when the script subtag is the special value Zzzz.

The spec should clarify if likely subtags are to be added and if yes, under which conditions likely subtags are added.

uloc_isRightToLeft behaviour for some inputs:

Locale	Likely subtags added	Locale for lookup	Result
en	Yes	en-Latn-US	left-to-right
en-GB	Yes	en-Latn-GB	left-to-right
en-ZZ	Yes	en-Latn-US	left-to-right
en-Latn	No	en-Latn	left-to-right
en-Arab	No	en-Arab	right-to-left
en-Zzzz	No	en-Zzzz	left-to-right
en-Zzzz-ZZ	No	en-Zzzz-ZZ	left-to-right
ar	Yes	ar-Arab-EG	right-to-left
ar-ZZ	Yes	ar-Arab-EG	right-to-left
ar-Zzzz	No	ar-Zzzz	left-to-right

Maybe the spec should be rewritten to follow the steps mentioned here: unicode-org/icu4x#3172 (comment)