Introduce ArrayWrapper?D, replacing ConstExpression?DArrayCopy

+ no explicit copy constructor of structs needed any more
+ removed assignment operator of structs that don't hold a const sType member (for those, we still need the assignment operator!)
+ simplified operator==() of structs (pre C++20)
+ resolved some 32bit warnings

samples/DynamicUniform/DynamicUniform.cpp

@@ -86,7 +86,7 @@ int main(int /*argc*/, char ** /*argv*/)
     model = glm::translate(model, glm::vec3(-1.5f, 1.5f, -1.5f));
     mvpcs[1] = clip * projection * view * model;
 
-    VkDeviceSize bufferSize = sizeof(glm::mat4x4);
+    vk::DeviceSize bufferSize = sizeof(glm::mat4x4);
     if (limits.minUniformBufferOffsetAlignment)
     {
       bufferSize = (bufferSize + limits.minUniformBufferOffsetAlignment - 1) & ~(limits.minUniformBufferOffsetAlignment - 1);

samples/PipelineCache/PipelineCache.cpp

@@ -127,7 +127,7 @@ int main(int /*argc*/, char ** /*argv*/)
     {
       // Determine cache size
       readCacheStream.seekg(0, readCacheStream.end);
-      startCacheSize = readCacheStream.tellg();
+      startCacheSize = vk::su::checked_cast<size_t>(readCacheStream.tellg());
       readCacheStream.seekg(0, readCacheStream.beg);
 
       // Allocate memory to hold the initial cache data

samples/utils/utils.hpp

@@ -47,7 +47,7 @@ namespace vk
     }
 
     template <class T>
-    void copyToDevice(vk::UniqueDevice const& device, vk::UniqueDeviceMemory const& memory, T const* pData, size_t count, size_t stride = sizeof(T))
+    void copyToDevice(vk::UniqueDevice const& device, vk::UniqueDeviceMemory const& memory, T const* pData, size_t count, vk::DeviceSize stride = sizeof(T))
     {
       assert(sizeof(T) <= stride);
       uint8_t* deviceData = static_cast<uint8_t*>(device->mapMemory(memory.get(), 0, count * stride));