Improve the performance of the flowupdate logic

The current implementation of the flow update logic in the Core uses a number of callbacks (triggered by the model after a dB update) when processing flows and in some cases serialize certains actions (e.g. writing to the dB and sending the flows to the adapters). While this works well, the entire flow update performance will degrade when there are more devices per olt in the network. We should therefore optimize this mechanism by minimizing the number of callbacks and let actions that can run in parallel do so.