Design/extend benchmark to cover QML object-creation compiler use cases

On top of my head:

• Design fairly simple set of QML types to be benchmarked (or borrow from existing benchmarks)
• Compare the startup times of generated C++ vs QQmlComponent (can expect faster startup of generated C++ due to no type erasure/QQmlElement<T> wrapping, etc.)

1. case 1: measure object construction time only
2. case 2: measure object construction + compilation unit loading for generated C++ (on-construction CU load can be triggered by "Component.onCompleted"?)

• Compare the memory footprint of generated C++ vs QQmlComponent (in theory generated C++ requires less memory, worth checking)