Support _Float32 in libm_alias_float.