🐞BUG: not working with 'chunked' data! (fix inside)

test with http://emojitracker.com/
It sends 'chunked' data with headers: ```"transfer-encoding":"chunked"```
To get this working we need three things:
1) catch the 'transfer-encoding' === 'chunked'
2) don't .push to resData but send to user
3) don't close the socket but write to it

You need to add code to catch chunked data :
from
```
...
res.on("data", function (chunk) {
resData.push(chunk);
}
...
```
to
```
...
res.on("data", function (chunk) {
if (resHeader['transfer-encoding'] === 'chunked') {
processResponse([chunk], {'chunked': true});
} else {
resData.push(chunk);
}
}
...
```

Then you need to move whole async code
```
...
var serverResData;
async.series([

//ungzip server res
function (callback) {
...
```
to the function
```
function processResponse(resData, chunked) {
var serverResData;
async.series([

//ungzip server res
function (callback) {
...
```

notice it takes resData as an argument!

and replace the code there for response not to close the socket
```
//send response
}, function (callback) {
if (global._throttle) {
console.log(21);
var thrStream = new Stream();

var readable = thrStream.pipe(global._throttle.throttle());
readable.pipe(userRes);

thrStream.emit("data", serverResData);
thrStream.emit("end");
callback();
} else {
userRes.end(serverResData);
callback();
}

```
to
```
//send response
}, function (callback) {
if (global._throttle) {
console.log(21);
var thrStream = new Stream();

var readable = thrStream.pipe(global._throttle.throttle());
readable.pipe(userRes);

thrStream.emit("data", serverResData);
thrStream.emit("end");
callback();
} else {
if (chunked) {
userRes.write(serverResData);
} else {
userRes.end(serverResData);
}
callback();
}
```

Start at the response handler containing res.on("data"), resData, processResponse, and the async.series response flow. Check the behavior with emojitracker.com or another transfer-encoding: chunked response, then verify that chunked data is forwarded incrementally without closing the socket while non-chunked responses retain their existing completion behavior.

Written by the indexing model from the issue text.