QTBUG-58083 asks for consistency in using the value of pi. One way to become consistent is to use C++20 std::numbers::pi or std::numbers::pi_v<T>. Since we can't depend on C++20, yet, it may be worthwhile to back-port this feature into q20.

Note that q20::numbers::pi_v<qfloat16> should also work, which means we need to specialize the variable templates for qfloat16 in qfloat16.h in either namespace std::numbers or q20::numbers, depending on __cpp_lib_math_constants.

q20::numbers::e_v<long double> is optional, but if it's absent, it must be documented as such:

// like std::numbers (except no support for long double)