be explict about softfloat's float128_t usage instead of boost::float128_t; fixes boost 1.80 build

[spoonincode]

abi_serializer.cpp does a

using namespace boost;

Apparently something internal to boost has been reorganized and when building with boost 1.80 beta it appears this pack_unpack<float128_t>() is getting confused with boost::float128_t (gets a template argument deduction/substitution failed error).

Be explicit here that we want softfloat's float128_t.

Sending this in to 3.1 since confusion around this type seems like a Bad Thing. I wonder if we should blanket replace all float128_t with ::float128_t elsewhere though?

[heifner]

I wonder if we should blanket replace all float128_t with ::float128_t elsewhere though?

I would vote, yes. If/when float128_t added to C++ it will likely be in namespace std. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2021/p1467r5.html . Seems like we would be better served to be explicit. Maybe even better to declare our own using soft_float128_t = ::float128_t

[spoonincode]

hm yeah let me think that over some

[spoonincode]

replaced by #702