Efficient implementation for `StrictSeq` operations

[lehins]

Many operations on StrictSeq simply rely on the underlying lazy Seq implementation that is simply followed by a forcing all of the elements.

I haven't checked, but I am sure most of the can be implemented in a more efficient manner.

For example Functor instance can do an fmap in one pass, instead of two, since Seq does have a Monad instance:

instance Functor StrictSeq where
  fmap f (StrictSeq s) = StrictSeq (f <$!> s)

Another example from a PR review:

This is a sub-optimal implementation IMHO, since forceElemsToWHNF and consequently forceToStrict cause a redundant iteration over the resulting data structure, just to force the elements. It would be beneficial to have custom implementation instead.

As far as this PR is concerned, there is no need to change anything. I'll turn this comment into a ticket instead that someone can work on and do proper testing and performance analysis.

scanl f z0 (StrictSeq xs) = 
  z0 <| StrictSeq $! (snd (mapAccumL (\ x z -> let !x' = f x z in (x', x')) z0 xs))

Originally posted by @lehins in #592 (comment)