chore: Improve locale affinity logic with LanguageUtils

[ericfjosne]

We can improve and optimize the locale affinity logic by making use of the newly introduced LanguageUtils class.

Instead of relying on a score-based affinity only, we first calculate a "same language" affinity by testing for corresponding spoken language.

This actually has improved the results for several locales:

• Locales for which the language code is sufficient to identity a language are now matched with an affinity SAME_OR_INTERCHANGEABLE. The score based implementation enforces script matching, which was not a good fit for the affinity logic. For instance: hr (Croatian) and bs (Bosnian) are now matched with each other with an affinity SAME_OR_INTERCHANGEABLE, whether Bosnian is written in Latin or Cyrillic script, which is a more accurate outcome.
• nn (Norwegian Nynorsk) is now matched with nb (Norwegian Bokmål) with an affinity SAME_OR_INTERCHANGEABLE instead of LOW which is a more accurate outcome.

It is also a nice logic optimization, as we won't systematically be calculating maximized locales (which is a costly operation).

[RichardNysater]

Nice optimization