multiprecision, correct code with cpp_rational doesn't compile

[Stepan Podoskin <stepik-777@…>]

This code gives me a lenghty error messaage:

#include <boost/multiprecision/cpp_int.hpp>

int main()
{
    boost::multiprecision::cpp_rational a;
    boost::multiprecision::cpp_int b;
    a = -b + a;
    return 0;
}

[John Maddock]

(In [83227]) Fix operators that operate on unary-negated expressions. Suppress some MSVC specific warnings. Fix some bugs exposed by above changes and add new test cases. Fixes #8176.