add GCC unreachability intrinstic on false branch of assert

[scarf005]

Summary

SUMMARY: Bugfixes "add GCC unreachability intrinstic on false branch of assert"

Purpose of change

/home/scarf/repo/cata/Cataclysm/src/vpart_position.h:52:20: error: ‘*(const vpart_position*)((char*)&<unnamed> + offsetof(vehicle_part_iterator<vehicle_part_range>,vehicle_part_iterator<vehicle_part_range>::vp_.cata::optional<vpart_reference>::<unnamed>)).vpart_position::part_index_’ may be used uninitialized [-Werror=maybe-uninitialized]
   52 |             return part_index_;
      |                    ^~~~~~~~~~~

• fix maybe-uninitialized compilation error
• fix from Insert GCC unreachability intrinsic in false branch of cata_assert when NDEBUG is enabled CleverRaven/Cataclysm-DDA#58210

Describe the solution

• redefine assert to use __builtin_unreachable() on GCC
• only applied on vpart_range.h since BN does not have cata_assert.h

Describe alternatives you've considered

• cherry-pick cata_assert.h

Testing

Compiles on GCC

[joveeater]

This shouldn't be needed. Assertions should work like this anyway. DDA needed to do this but only because they use their own implementation of assertions so they had to mimic this behavior.

The error message that you gave, i.e. vpart_position.h line 52, doesn't have an assertion in that function so it seems very weird that this fixes that problem. More importantly this could trip people up. It means including a seemingly innocuous vehicle parts header causes assertions to behave differently and not get checked even on the debug versions.

Redefining assert should be a last resort. I would try giving the property a default value first. Hopefully that'll shut gcc up, if not I'd just go with the pragmas until they sort out their junk warnings.