long double support in QDataStream

Add support for streaming long double into and out of QDataStream. This has to be in a portable format. long double may be 64bits, 80bits or 128bits, depending on the platform, so serialization should be as a 128bit BE value. It's trivial to add when sizeof(long double)*8 == 128. On other platforms, we need to convert into a 128bit quantity w/o stdlib help.

• Clang and GCC have __float128, but probably not on all platforms?
• Boost.Multiprecision has a float128 type, but only on GCC and ICC: https://www.boost.org/doc/libs/1_81_0/libs/multiprecision/doc/html/boost_multiprecision/tut/floats.html
• ...?