Improved middleware runner performance

[WyriHaximus]

This middleware runner is faster and uses less RAM the more middleware added because it doesn't creates new instances of it self while iterating over all the middleware.

[jsor]

Not strictly related to this PR, but the runner could need a test that it can be run multiple times without internally modifying the middleware stack. This might be avoiding regressions when refactoring.

$runner = new MiddlewareRunner([
    $this->expectCallableExactly(2),
    $this->expectCallableExactly(2)
]);

$runner(new ServerRequest('GET', 'https://example.com/'));
$runner(new ServerRequest('GET', 'https://example.com/'));

[WyriHaximus]

@jsor could you PR that, so we can add that test before merging this PR?

[jsor]

@WyriHaximus damn, i failed trying to offload work 😄 Here it is: #237

[WyriHaximus]

@jsor You can try 😝 . Cheers 👍

[WyriHaximus]

Ping @clue @jsor rebased from master and fixed the middleware runner to pass all tests

[clue]

The test is currently failing, otherwise LGTM 👍

Can you give some numbers on how this is "improved"?

[WyriHaximus]

@clue yeah only 5.3 test is failing at the moment. Current master is more over 3 times slower than this PR. Also only only has all the middlewares in memory once and no extra references to them like the current master has saving memory. (Depending on how many middlewares you have and how many requests you're handling at that moment that adds up.)

[WyriHaximus]

@jsor @clue fixed PHP 5.3 support and it seems that doubled the performance

[jsor]

IMHO, this should be $response insteand of $cancel.

[WyriHaximus]

@jsor Good point, changed it 👍

[clue]

Nice, let's get this in! 📈 🎉