vulkan: Add workaround for device free memory > total memory on Intel Windows (#29835)

* add workaround for Intel Windows driver issue

* applying safeguard to all GPUs

also a minor refactor

* partial revert

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -15265,7 +15265,10 @@ void ggml_backend_vk_get_device_memory(int device, size_t * free, size_t * total
             *total += heap.size;
 
             if (membudget_supported && i < budgetprops.heapUsage.size()) {
-                *free += budgetprops.heapBudget[i] - budgetprops.heapUsage[i];
+                // Driver may report heapBudget > heap.size so we add a safeguard here
+                // See https://github.com/ggml-org/llama.cpp/issues/29277
+                const vk::DeviceSize budget = std::min(budgetprops.heapBudget[i], heap.size);
+                *free += budget > budgetprops.heapUsage[i] ? budget - budgetprops.heapUsage[i] : 0;
             } else {
                 *free += heap.size;
             }