GCC: Investigate and fix what's missing for _Float128 support

[vejbomar]

Windows-on-ARM-Experiments/gcc-woarm64#3 changes definition of long double from 128bit to 64bit to be aligned with Clang and MSVC for aarch64 on Windows. After that 128bit float number is available as _Float128 but not everything that is working on aarch64-linux-gnu is working on aarch64-w64-minw32

What works:

• _Float128 type exists, it can be used to define variables, in function args etc.
• suffix Q works for defining 128bit numbers, e.g. _Float128 e_contant = 2.718281828459045235360287471352662498Q

What doesn't work (not sure if all should work but these mostly work with aarch64-linux-gnu target):

• Suffix F128 doesn't work for defining 128bit numbers
• Q length modifier in printf doesn't work (afaik this is gcc's extension, standard doesn't specify this: https://en.cppreference.com/w/c/io/fprintf, page 288 of https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf)
• strfromf128 standard function for converting 128bit floats to string isn't available
• std::cout << 2.7182Q << std::endl doesn't work

Sources: https://cpufun.substack.com/p/portable-support-for-128b-floats
Older branch which was enabling _Float128: https://github.com/Windows-on-ARM-Experiments/gcc-woarm64/tree/float128

[vejbomar]

After enabling support for _Float128, changes from Windows-on-ARM-Experiments/gcc-woarm64@4979060 and Windows-on-ARM-Experiments/gcc-woarm64@584885c which are part of 64bit long double PR, should be probably reverted