Offer safe variants for the RIO.List.Partial functions in the RIO.List module

[roman]

Iterating over the work from #43, what do you think about adding all the functions from RIO.List.Partial with a safe prefix in the RIO.List module?

There is no clear replacement for the functions in RIO.List.Partial, and I would avoid them as much as possible if I can.

[snoyberg]

In principle: I'm OK with this approach. I'm not convinced that safe will be the right prefix in all cases. For example: maximum would probably be better replaced by maximumMay or maximumMaybe to match existing naming conventions IMO.

[roman]

That is fair, honestly, my intent was not to strongly advocate names, may I create a pull request and we revisit the names as we go?

[mitchellwrosen]

FYI: scanl1, scanr1 aren't partial

[snoyberg]

Implemented in #96

[akhra]

Late entry, but perhaps relevant... yesterday I had to switch some lists over to vectors so I could use statistics operations on them. I had been using some of the (list-specific) *Def functions from safe, and needed to replicate those; and some of the statistics operations I wanted were partial on empty vectors, too, which needed fixing. I wound up writing this:

defaulting :: Foldable f => b -> (f a -> b) -> f a -> b
defaulting d f xs
  | null xs   = d
  | otherwise = f xs

So now I have e.g. defaulting 0 Vector.maximum. The *Maybe pattern is a trivial special case:

mayhap :: Foldable f => (f a -> b) -> f a -> Maybe b
mayhap f = defaulting Nothing (Just . f)

Are there any particular downsides to this technique? It seems far more elegant than a slew of specialized, type-specific functions.

Edit: https://hackage.haskell.org/package/safe-foldable