SPIKE: Research QMetaObjectBuilder efficiency

The QML engine creates many dynamic meta-objects, and if we reduce the usage of property caches, this usage might even increase. Thus QMetaObjectBuilder should be as efficient as possible.
Especially when it comes to meta-methods, there should still be a few low-hanging fruits.

This task is about figuring out:

• Can we get a benchmark for the builder?
• Where do we do avoidable work in the builder?
• Can we make use of the improved metatype system in it (in more places)?