Potentially more efficient `fromMap` operation.

[jonathanknowles]

The fromMap operation is currently defined as:

fromMap :: MonoidNull v => Map k v -> MonoidMap k v
fromMap = MonoidMap . Map.mapMaybe maybeNonNull

Internally, this always creates a new Map structure, even if the original Map is already in canonical form (i.e., contains no mempty values).

It might be desirable to first check whether the provided Map is in canonical form, and if so, just perform a coercion, thus avoiding the creation of a new Map structure:

fromMap :: MonoidNull v => Map k v -> MonoidMap k v
fromMap m
    | not (any MonoidNull.null m) = coerce m
    | otherwise = MonoidMap (Map.mapMaybe maybeNonNull m)

However, one downside of this optimisation is that it penalises the general case, which would now require two traversals of the original map, whereas before only one traversal was required.

We might be able to get the best of both worlds by providing two operations, and allowing the caller to choose:

fromMap :: MonoidNull v => Map k v -> MonoidMap k v
fromMap = MonoidMap . Map.mapMaybe maybeNonNull

fromMapNonNull :: MonoidNull v => Map k v -> MonoidMap k v
fromMapNonNull m
    | not (any MonoidNull.null m) = coerce m
    | otherwise = fromMap m