fix: setLocaleFromURL handle fallback locals and regex of BCP-47 Script/Region 🌋

[darcywong00]

Found while integrating Keyman for Android which reports Spanish locale as es-ES

Had to make the following changes to setLocaleFromURL():

• check with fallback locales

• Also update BCP-47 regex to capitalize Script and Region

This way, the following link shows in Spanish (which uses es.php files)
http://keyman.com.localhost/es-ES/keyboards/?embed=android&embed_version=19.0

invalid locale like
http://keyman.com.localhost/rv-tS/keyboards/?embed=android&embed_version=19.0

get a page not found.

I couldn't invalidate the locale because that broke _test pages like /_test/go (would treat _test as an invalid locale)

Test-bot: skip

[keymanapp-test-bot]

User Test Results

Test specification and instructions

User tests are not required

[darcywong00]

force rebased onto staging

yeah, we can rebase to staging and close the web-i18n epic branch, so that we don't have confusion. When ready, we just make a back-port to master from staging PR

[mcdurdin]

I think we want our URLs to redirect if a fallback is encountered, rather than accepting es-Fooo-BA as a substitute for es. And we don't want case insensitivity either; we don't want es, eS, Es, and ES all to return the same document, because URLs are case sensitive. So ... maybe we need to take this one back to the drawing board -- happy to discuss further.