Avoid trying to free null RIDs in FSR2 teardown

[RandomShaper]

On shutdown, removes nine of these from happening:

ERROR: Attempted to free invalid ID: 0 
   at: RenderingDeviceVulkan::_free_internal (drivers\vulkan\rendering_device_vulkan.cpp:8557)
drivers\vulkan\rendering_device_vulkan.cpp:8557 - Attempted to free invalid ID: 0 

CC @DarioSamo

[DarioSamo]

Seems good to me, I guess I figured free() works on null like how the C one does.

[RandomShaper]

As far as I can tell it's a common pattern in the rendering code to check if an RID is valid (non-null) and, in more complex areas of the code, even if a non-null RID hasn't been already released somewhere else (because of cascade freeing that RD can do). Having the null check before every free is maybe silly, when free() could already have the C's free()-like semantics you mention. Food for thought and maybe digestion for improvement. 😃 (I won't get to the next step after digestion. 🤣)

[lawnjelly]

We had the same thing in 3.x .. VisualServer had a safe check for invalid RIDS until #55764 (author was very keen for this but this seemed like PTSD, and I'm not sure it has helped identify any bugs 😀 ).

This has predictably caused some regressions with Dummy servers which return NULL RIDs, and just meant client code all needs this check instead. On reflection I think it was probably better for servers to silently ignore NULL RIDs, as does ::free() etc and most library funcs.

[YuriSizov]

Thanks!