LSP violation

[mapogolions]

Suppose we have a function f that perfectly works with some domain of values called A.
[Domain A/input] -> f -> [Codomain A/output]
At the same time, we have another function g that works fine with another domain of values called B.
[Domain B/input] -> g -> [Codomain B/output]
The function f can safely replace the function g if and only if the domain B is a subset of the domain A && the codomain A is a subset of the codomain B.
Let's look at the following signatures

• Headers::setHeader(string $name, ...) - f(domain A is string type)
• Message::withHeader($name, ...) - g(domain B is any type)

When we do something like this $response->withHeader($any, ...) we violate the above principle. The specified method call is transformed to $this->headers->setHeader($any, ...).
The domain A (string type) is less than the domain B (any type). For this reason, unnecessary static methods had to be implemented causing a problem.

[coveralls]

Coverage decreased (-0.001%) to 99.73% when pulling 5f8a1fb on mapogolions:enhancement/LSP-violation into 0b3be28 on slimphp:master.

[l0gicgate]

I think we may have gotten carried away with the idea of strong typing every method as much as possible. I agree that we are violating said principle. @adriansuter I'm fine with what is being proposed. Please review before I merge.