Add HUGE_VAL macros for more _FloatN, _FloatNx types.

This patch continues the preparation for additional _FloatN / _FloatNx
type support by adding appropriately conditional definitions of
HUGE_VAL_* macros, corresponding to the HUGE_VAL_F128 definition
already present.

Tested for x86_64.

	* math/math.h [__HAVE_FLOAT16 && __GLIBC_USE
	(IEC_60559_TYPES_EXT)] (HUGE_VAL_F16): New macro.
	[__HAVE_FLOAT32 && __GLIBC_USE (IEC_60559_TYPES_EXT)]
	(HUGE_VAL_F32): Likewise.
	[__HAVE_FLOAT64 && __GLIBC_USE (IEC_60559_TYPES_EXT)]
	(HUGE_VAL_F64): Likewise.
	[__HAVE_FLOAT32X && __GLIBC_USE (IEC_60559_TYPES_EXT)]
	(HUGE_VAL_F32X): Likewise.
	[__HAVE_FLOAT64X && __GLIBC_USE (IEC_60559_TYPES_EXT)]
	(HUGE_VAL_F64X): Likewise.
	[__HAVE_FLOAT128X && __GLIBC_USE (IEC_60559_TYPES_EXT)]
	(HUGE_VAL_F128X): Likewise.