octotiger: add boost_multiprecision and cxx20 variants

[G-071]

Thanks, LGTM! Though I will likely modify the cxx20 variant into some like cxxstd=[17,20] to better accommodate for future C++ standards. I can also add an variant which completely disables quadmath if ever required as it is only needed for a subset of our tests (though as long as the boost_multiprecision variant works there's probably little reason to do so).

[junghans]

You will have have to hand the cxxstd down to hpx and boost.