[inplace_function] Should [](){return 42;} be convertible to inplace_function<void()>? about sg14

[Quuxplusone]

Survey of existing practice (that is, of people who provide a single-header implementation, because I'm lazy): https://godbolt.org/z/PGx6JA

std::function happily permits converting int(*)() to function<void()>.

boost::function happily permits converting int(*)() to function<void()>.

https://github.com/TartanLlama/function_ref tl::function_ref consistently forbids converting int(*)() to function<void()>.

https://github.com/Naios/function2 fu2::function claims to permit converting int(*)() to function<void()>, but if you try it, you get a hard error from the guts of the vtable.

After #155, inplace_function claims to permit converting int(*)() to function<void()>, but if you try it, you get a hard error from the guts of the vtable.

I'd be interested to learn what happens with folly::function; with HPX unique_function; and/or with MongoDB unique_function. But they're not single-header Godbolt-friendly, so I didn't really try hard.