Disabling Float128 does not work

[kursatyurt]

In the develop branch disabling float128 flag does not work.

In my cmake file, I added flag

add_definitions(-DBOOST_MATH_DISABLE_FLOAT128)

I can compile on 1.86 without any problem; however, it does not compile when I switch to the development branch

third_party/boost_math/include/boost/math/tools/big_constant.hpp(52): error: "T (boost::math::tools::largest_float, const char *, const std::true_type &, const std::false_type &)" contains a 128-bit floating-point, which is not supported in device code

The latest commit I can compile is 07d97df50 if I traced it correctly.

[mborland]

Fixed by: #1207