graphics: count DCC/CMASK metadata readbacks in the performance overlay

src/graphics/host_gpu/renderer/cache/textureCache.cpp

@@ -1070,6 +1070,9 @@ void TextureCache::MaterializeColorClear(ImageId id, const ImageDesc& desc,
 	// Finish native metadata writes before reading backing bytes. This can submit the scheduler,
 	// so discovery runs before final draw uploads and never holds the texture lock across it.
 	if (m_buffer_cache.IsRegionGpuModified(range.address, range.size)) {
+		Common::DebugCounters::ScopedTimer timer(
+		    Common::DebugCounters::Counter::ColorMetadataReadbacks,
+		    Common::DebugCounters::Counter::ColorMetadataReadbackNs);
 		m_buffer_cache.ReadMemory(range.address, range.size, false);
 	}
 	const auto slice_size = range.size / layers;

src/graphics/presentation/performanceOverlay.cpp

@@ -310,6 +310,9 @@ void GpuActivityRows(const Snapshot& shown) {
 	Row("CPU waiting on GPU",
 	    Fmt("%.0f ms/s  (%.0f waits/s)", Rate(shown, Counter::GpuWaitNs) / 1e6,
 	        Rate(shown, Counter::GpuWaits)));
+	Row("DCC/CMASK readbacks",
+	    Fmt("%.0f ms/s  (%.0f/s)", Rate(shown, Counter::ColorMetadataReadbackNs) / 1e6,
+	        Rate(shown, Counter::ColorMetadataReadbacks)));
 }
 
 void CompilationRows(const Snapshot& shown) {