[Do Not Merge] ReqBody support

[nsaunders]

This adds ReqBody support in probably the most naïve way possible. Related to #1, but offers no "notion of delayed IO".

[maxhallinan]

Thanks for opening this PR! What do we need to do to get this merged?

[nsaunders]

Good question.

If you look closely, this is a pretty crude implementation, definitely not production-ready. The main issue is that the request body is read to a string on every request whether it is used or not. I haven't found a straightforward way around this yet, and I think significant design changes (probably even in Hyper) might be required to fix it.

@owickstrom has explained to me that Hyper and Hypertrout are considered experimental, so he may be OK with integrating something like this with the understanding that it's really only there to demonstrate a concept; but I would definitely not disagree with a decision to wait until a better solution can be found.

In the meantime, you can feel free to clone my fork and experiment with my reqbody branch (the one you see in the PR).

[nsaunders]

@maxhallinan I have also been working on Nodetrout, which is more feature-complete than Hypertrout. (As an implementation detail, it sacrifices the type safety of Hyper, but, for me, it is a reasonable tradeoff.)