Add isNull predicate

[anttih]

Allows writing a faster decoder for optional values because you don't have to call case_ twice to first peel off the null case and then do the actual decoding for the value. This came up while porting backend-optimizer to use json where I had to add a nullable:

nullable :: JSON -> Maybe JSON
nullable json = case_
  (const Nothing)
  (Just <<< fromBoolean)
  (Just <<< fromNumber)
  (Just <<< fromString)
  (Just <<< fromJArray)
  (Just <<< fromJObject)
  json

I can add predicates for other types as well if needed.

Ping @natefaubion

[garyb]

Sounds good to me. I think we can do without the other predicates - I was aiming to keep the FFI footprint of this library as small as possible for portability reasons - but null does seem like a special case where you'll often want to guard against it rather than fold over the all the available types. I guess there is toNull too which kinda works for that, but it's not really like the others since you're only going to care about the Just-iness of the reuslt.

[natefaubion]

It's not totally obvious to me why this needed to be FFI, FWIW. a toNull implemented with case_ is OK enough. For optimization purposes I just need a function to exist in the library to hook into with obvious semantics that can potentially be inlined during codegen, but it doesn't necessarily need to be FFI.