Add take, drop, and splitAt

[treeowl]

The least we can do is use fromListN machinery to make take linear in the result size, which smells big-O optimal. For stack drop, I imagine we can do something more interesting, to make it linear in the number of elements dropped. For queues, I've implemented take in time linear in the number of elements taken, but I suspect we can do better than that in the case where we take most of the elements—some sort of "drop from the rear" smells better.

• take for stacks—completed in Implement take for stacks #15
• take for queues—implemented in Implement take for queues #17, but I suspect we can improve it.
• drop for stacks
• drop for queues
• splitAt for stacks
• splitAt for queues