product_type_base Compared to std::tuple is inconsistent

[mochunshuang]

old: -> S

template <::std::size_t J, typename S>
   static auto element_get(::beman::execution26::detail::product_type_element<J, S>&& self) noexcept -> S {
       return ::std::move(self.value);
   }

new -> S &&

template <::std::size_t J, typename S>
   static auto element_get(::beman::execution26::detail::product_type_element<J, S>&& self) noexcept  -> S && {
       return ::std::move(self.value);
   }

because

  std::tuple<int, double, std::string> t = {42, 3.14,
                                               std::string("Hello, World!")};
     static_assert(std::is_same_v<decltype(std::get<1>(std::move(t))), double &&>,
                   "get<1> should return double&&"); 

auto t = product_type{ //old version
           42, 3.14, std::string("Hello, World!")};
  static_assert(not std::is_same_v<decltype(std::move(t).get<1>()), double &&>,
                     "get<1> should return double&&");// where inconsistent

[dietmarkuehl]

I vaguely recall that this is what I had. Unfortunately, some analysers don't like get<I>(std::move(pt))... where I is parameter pack: they think that pt is used multiple times. Maybe the fix is to just reformulate it the other way around. I'll have a look.

[dietmarkuehl]

I have made the change on this PR. It turns out the problem was a different one - the test actually tested the wrong thing (the reference qualification was on the wrong template argument).