[GPU] Converted Depth/Color Surfaces to a ring buffer

Additionally made the surface cache search hit for any address that lies in the surface.

This should allow multiple races to be done in Mario Kart and fixes the intro video.

include/renderer_gl/surface_cache.hpp

@@ -46,7 +46,7 @@ public:
 
     OptionalRef findFromAddress(u32 address) {
         for (auto& e : buffer) {
-            if (e.location == address && e.valid)
+            if (e.location <= address && e.location+e.sizeInBytes() > address && e.valid)
                 return e;
         }