As of today all answer seem to be correct: No way to do it.
I would like to add some meat to those statements:
Wikipedia on "automatic parallelization".
Automatic parallelization would be the process to take a given serial algorithm and try to find command sequences that could be executed in parallel, then scheduling and dispatching those sequences to the available (hardware: cpu) resources, and then joining the outcome (=data) back for further processing.
The problem not only is in finding sequences that do not depend on each other and are therefore able to being worked on in parallel. But also the communication overhead may become prohibitive. Think about your particular problem where the one Minecraft world is going to be split in some way onto the different cpus: If you split the world geometrically, lets say in := 9 blocks, then at the borders the individual sub worlds need to exchange any world modifications or movements of inhabitants.
While the above is a typical "divide-and-conquer" approach, and geometrical split-up of a problem is a widespread solution to parallelization, it is normally handcoded, making use of the knowledge the problem designer - in this case - the game developer has.
Your proposed idea would involve some automation of an algorithm that is completely unknown, because the virtual "hyper" server has no way of knowing how Minecraft internally behaves. So the only angle to this problem is the executed assembly code. At this low a level it is very hard - and currently seems to be impossible - to come up with a solution that has any merit.
Google can search for scientific publications these days. And if you're really interested I'd look at the publications on related conferences. Organizations such as the IEEE and the ACM may be good starting points for that.