cardano-data cleanup

[goolord]

• removes some usages of fromSJust
• move encodeKeyedStrictMaybe to cardano-data
• Introduce ofield and invalidField. I'm not totally sold on these actually, but they're both common patterns and it would be nice if the api nudged you towards writing the SparseKeyed fields in the way it wants you to. A really easy mistake I'm trying to prevent is field (\x foo -> foo { fooX = x }) (From :: Decoder ('Closed d) (StrictMaybe X)) where x is actually omittable
• Introduce MapE
• Split up the definite and indefinite length branches of getSparseBlock, into separate functions should very mildly reduce allocations and help branch prediction

[lehins]

I don't have enough familiarity with coders package to say anything definitely about the correctness of the PR, but at a high level it looks great. I've requested @TimSheard to review it, since he'll probably be the best judge in this area.

[JaredCorduan]

LGTM (but always happy to have another review, if Tim still wants to have a look)