Need `long double` (float128) routines support

[AngheloAlf]

GCC can emit calls to these routines when using the n32 ABI when there's arithmetic on long double.

Type definitions:

typedef long double float128;
typedef int fcmp __attribute__ ((mode (__libgcc_cmp_return__)));

Required functions:

• float128 __extenddftf2(float64 a);: float64 -> float128. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__extenddftf2
• int32_t __fixtfsi(float128 a);: float128 -> int32_t. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__fixtfsi
• float128 __floatsitf(int32_t a);: int32_t -> float128. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__floatsitf
• float128 __subtf3(float128 a, float128 b);: subtract. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__subtf3
• float128 __multf3(float128 a, float128 b);: multiply. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__multf3
• float128 __divtf3(float128 a, float128 b);: divide. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__divtf3
• fcmp __gttf2(float128 a, float128 b);: >. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__gttf2
• fcmp __lttf2(float128 a, float128 b);: <. https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__lttf2