Change the lookup function to be a better fallback chain

[valkum]

Closes #1

Instead of using the computed fallback chain, lookup_complete for the two loaders will now use an ordered list of matching languages. This allows to overwrite translations for languageIds with higher specifier. e.g. have a baseset in en, and differences for en-GB in en-GB

[XAMPPRocky]

Thank you for your PR! This looks good to me, but it looks like it's failing the tests.

[valkum]

Sry for the delay.

So basically going for the proposed solution from #1 this results in the fallback key to be present in the base zh language.

I am not sure if we should change that to have a resolution order like: zh-CN -> zh-CN, zh-TW, zh, default_fallback

The question then would be: What to do with a specificity higher than 2? de-DE-1996 -> de-DE-1996, [de-CH-VARIANT], de-DE, de-CH, de, default_fallback? Not sure how the best approach should be. If we consider the current state as a linked list, other language regions like CH would then be siblings in a tree. The question is how many branches of the main path between the requested one and the root should be visited.

fallback <- de <- de-DE <- de-DE-1996
               Î         ^- de-DE-VARIANT2          
               \ de-CH <- de-CH-VARIANT

In which order should they be visited?

And what about multiple requested languages? (possibly with weights), can we ignore them for now? I guess currently one can argue that this is only localization resolution and language negotiation should be done ahead.

[XAMPPRocky]

cc @zbraniecki Would you happen to have advice / opinion on how to handle the above?

[XAMPPRocky]

Thank you for your PR, and congrats on your first contribution! 🎉

I'm going to merge this as is and open a new issue for the other fallback chain issues.