<numeric>: non conformance of std::reduce()

Describe the bug
in std::reduce(), binary_op(*first,*first) must be convertible to T. This isn't the case in the below code yet it compiles.

Command-line test case

C:\Temp>type repro.cpp
#include <iostream>

int main() {
  std::random_device e;
        std::uniform_int_distribution<> dist(1, 10);
        const int n = 10;
        std::vector<int> v(n);
        std::generate(v.begin(), v.end(), [&]() {return dist(e); });  
        const auto result = std::reduce(v.begin(), v.end(), std::make_pair(0, 0),
            [](std::pair<int,int> sum,int n) {
                n % 2 == 1?sum.first += n:sum.second += n;
                return sum;
            });
}

C:\Temp>cl /EHsc /W4 /WX .\repro.cpp
Microsoft (R) C/C++ Optimizing Compiler Version 19.26.28806 for x86
Copyright (C) Microsoft Corporation.  All rights reserved.

repro.cpp
Microsoft (R) Incremental Linker Version 14.26.28806.0
Copyright (C) Microsoft Corporation.  All rights reserved.

/out:repo.exe
repro.obj

Expected behavior
It shouldn't compile

STL version
Microsoft Visual Studio Community 2019 Version 16.6.1

Additional context
gcc and clang reject the above code

Start with the provided repro.cpp and the std::reduce entry point; compare MSVC's acceptance with the stated gcc and clang behavior. Done means the invalid binary operation is rejected, with a regression test covering this case.