Optimize read/write operations between models

[davidhjp01]

With large numbers of FMU's input and output ports and connections, there is a significant performance degradation due to many iterative value transfers (reads/writes) between FMU ports.

I have only briefly went over how this operation is performed. But for the simulations without using ProxyFMU, we can probably optimize this process via pointer swapping? Also there are value modifiers, which we can use of some parallel loops (any other suggestions?) .

[THEvang]

@restenb and @davidhjp01 will do profiling to pinpoint the performance bottleneck