Historically we weren't very disciplined in how we passed around a Lang on which perform corpus search. Not only that, but our Lang type had (before this MR) an All data constructor which morally didn't belong to the Lang type, which is clearly a bounded enumeration of all the languages that we support for corpus search and stemming. Furthermore, these days the frontend is always going to send us a valid Lang to begin with, i.e. if the ExternalAPIs provider doesn't support searching for a language or the user didn't pick a language to filter the corpus to begin with, the frontend will default to EN. If we really don't care about the language the provider should search for, we can always just use Maybe Lang in the relevant API function for the given provider.

Removing the All data constructor from the Lang datatype has several benifits, the main one being that we now have total functions for things like toISO693.

This will make possible to write an uniform interface for stemming stem :: Lang -> StemmingAlgorithm -> T.Text -> T.Text, as now Lang will always map to a valid (ISO693) language.

While this could technically be merged, it's a stepping stone for #324 (closed).