Re-evaluate sandbox memory model

We should reconsider the requirements we have for the Hyperlight Sandbox and evaluate the sandbox memory model we have now, to decide whether it needs a complete redesign or small changes.

The main incentive for this discussion is the introduction of the `mmap`ed memory, that at the moment, is treated separately, outside the Shared Memory that ensures exclusive access.

We would also benefit from a simplification of the model, if possible.