enhancementhelp wanted
Description
I run a simple benchmark calling window API's GetSystemTime using JavaCpp's built-in windows API wrappers. This code allocates a struct, calls the native API and fetches some field from the struct:
SYSTEMTIME systemtime = new SYSTEMTIME();
windows.GetSystemTime(systemtime);
return systemtime.wSecond();
Profiling shows that the first line takes >90% of the overall execution time

I believe that there is some space for optimization here. The same thing implemented with Bridj or JNR outperforms JNI+JavaCpp just because of faster allocation, see the benchmark at https://github.com/zakgof/java-native-benchmark.
Say, with Bridj allocation takes <50% of the overall time:
