Remove our useless atEnd handler, since it was interim to begin with.…

… Protip: "this is not correct" in a comment means it is very likely to contain a bug :)

index.js

@@ -68,6 +68,10 @@ module.exports = {
            //
            // If we forget to call next(), our stream will just never end
            // and things may hang.
            //
            // The second function is optional, used if we need to do anything at
            // the end of the stream. We don't, since we just pipe any remainder
            // on blindly to our caller.

            // For simplicity, we're going to _completely_ ignore that binary
            // data like images are a thing that could ever exist.
@@ -151,24 +155,6 @@ module.exports = {
            }

            next(); // Ready for next chunk.
        }, function atEnd(next) {
            // through2 takes 2 functions: forEachChunk and atEnd -- this is the second one,
            // it gets called when the data incoming to through2 ends. Any clean-up or final
            // actions we need to take before we stop getting called happen here.

            // there's still a stream 'end' event that has to happen, after this stream will
            // do no more. We don't want a race between our cleanup and the caller's last read
            // or the end event. They may assume they got it all if end was emitted before this,
            // but we had one more thing to do, so we have to tell through2 we're not just done,
            // but done-done.
            //
            // If we had no atEnd, it would just send 'end' when we're done. But we had one thing
            // pending.

            // So if we had leftovers, let's put them in headers. This isn't correct, but
            // vanishing data in a closure variable is no good either.
            response.rawHeaders.push(partial);
            next();
        })).pipe(response);
        // Whatever comes out of our header-splitting stream parser must be the
        // body. Because that's what we designed, right?