We currently reject _ as a subtag separator when parsing a locale.

This has come up in unicode-org/icu4x#3336 .

I'm questioning the value of rejecting _ as a subtag separator on the input in favor of following the Robustness Principle.

We already do not follow Unicode BCP47 Locale strictly - we accept uncanonicalized Locale, for example this works:

new Intl.Locale("EN-fr");
new Intl.DateTimeFormat("aR-Pl");

both of those work despite not being canonical, we will canonicalize them on the output!

I suggest we extend support to:

new Intl.Locale("EN_fr");
new Intl.DateTimeFormat("aR_Pl");