GUI2/Dispatcher: refactored [dis]connect_signal to use constexpr-if

[Vultraz]

Opening as a PR for comments on the design, as well as waiting until we get the MacOS requirements sorted. Not 100% this is the correct use of std::is_convertible, but it does seem to work.

[stevecotton]

The error messages generated by both GCC and Clang are uninformative - they only hint at which .cpp file has the failing call:

In file included from src/actions/advancement.cpp:29:
In file included from src/gui/dialogs/unit_advance.hpp:16:
In file included from src/gui/dialogs/modal_dialog.hpp:18:
In file included from src/gui/core/static_registry.hpp:18:
In file included from src/gui/core/window_builder.hpp:19:
In file included from src/gui/widgets/grid.hpp:17:
In file included from src/gui/widgets/widget.hpp:17:
src/gui/core/event/dispatcher.hpp:526:4: error: static_assert failed "No matching signal queue found for event"
                        static_assert(false, "No matching signal queue found for event");
                        ^             ~~~~~
src/gui/core/event/dispatcher.hpp:570:4: error: static_assert failed "No matching signal queue found for event"
                        static_assert(false, "No matching signal queue found for event");
                        ^             ~~~~~

[stevecotton]

Opening as a PR for comments on the design

Will this check the callback function's signature at compile time, or will std::bind() manipulate the signature to fit and leave the error to be detected at runtime?

[Vultraz]

For reference, this is why I couldn't use static_assert(false): https://devblogs.microsoft.com/oldnewthing/20200311-00/?p=103553

[stevecotton]

Has this replaced a compile-time check with a runtime check?

[stevecotton]

For reference, this is why I couldn't use static_assert(false): https://devblogs.microsoft.com/oldnewthing/20200311-00/?p=103553

https://en.cppreference.com/w/cpp/language/if includes an example of how to do it as a compile-time check, using dependent_false_v.