Add historical currency codes

[MaxGabriel]

Closes #3

This is a draft attempt at closing #3, modeling historical currency codes

I say draft because:

1. There are 3 codes without numeric codes. I assigned 5000, 5001, and 5002 to them as placeholders. Those currencies could be removed from the list to simplify things
2. Many historical currencies have unknown numbers of decimal units. I defaulted those to 0. Possibly a -1 sentinel value could be used, or those currencies could be dropped entirely? I normally would try to model things faithfully, but correctly modeling would be a breaking change that only serves to model super old currencies.
3. I need to carefully review what is in the historical list or not against the ISO standard.
4. (Minor) I need to fix up the sorting, but holding off on that until the other stuff has a path forward

One alternative to trying to add a lot of historic currencies is to just add some on as-needed basis. For example, recent changes like HRK -> EUR are probably useful to people who have serialized HRK data, but it's less likely people need CSJ from 1953.

Any feelings on the above?

[isomorpheme]

Hmm - on the one hand, I personally don't mind the sentinel values much, as I don't realistically see them tripping up existing code. On the other hand, if we stick to this design, and we do want to model things differently after all, that would be a pretty painful breaking change.

My idea would be: throw more types at it, and introduce a separate type for historical currencies (and currency codes?). Then we can just use Maybe where values might be unknown, and users can make a deliberate choice about whether and how to handle those; for the most flexible case we can produce a type isomorphic to Either HistoricalCurrency Currency. This would just be an addition, so not a breaking change right now; moving a currency to historical would be breaking, but arguably that's a good thing, because then users are made aware that reality has changed (assuming it would be breaking in terms of compiler errors, not at runtime).

[isomorpheme]

I do realise that the question of which historical currencies one might want to handle still stands though: having to deal with a whole separate type for all historical currencies is probably annoying if you just want to handle recent cases like HRK.