Implement mkParsec—a new method of MonadParsec

[brandonchinn178]

Close #366.

Implements a new primitive as discussed in #366 (comment), which enables defining new primitives, at the cost of needing to drop (a little) into megaparsec internals.

[Lev135]

Why not to use Reply? It seems that your type of liftConsumer is isomorphic to

(State s e -> Reply s e a) -> m a

and implementation of pLiftConsumer is identical to mkPT.

[brandonchinn178]

I wasn't aware of Reply :)

1. Seems like mkPT would have to be exported
2. The docs of Reply says it represents the "result at the end of parsing". If we reuse it for this, we'd probably have to tweak the wording there?
3. Overall, not opposed. I'm just not sure how Reply is used currently, and if it should be reused here

[mrkkrp]

@brandonchinn178 So, first of all, I'm sorry it took so long to get to this one. This is a great idea which I believe can greatly empower users of the library. Regarding @Lev135's suggestion, I think it is a very good one. I sketched an alternative implementation and pushed it to this branch (I also took the liberty of rebasing on current master). Can you please check that the mkParsec method is sufficient for doing whatever you did with liftConsumer?

I did not implement breakOn just yet, but I you deem mkParsec usable we can add the example with breakOn back and merge this.

[brandonchinn178]

Looks great! Yeah, if you get the breakOn test working, that's sufficient for me!