Main (allocation) network

[SouthEndMusic]

In the discussions about allocation we have introduced the concepts of main network and subnetworks. In my current picture of this, the main network is disjoint from the subnetworks. As discussed, we could say that the main network as a special case of subnetwork denoted with subnetwork* ID 1, which is treated differently in the computational core than the other subnetworks.

Special properties of the main network:

• The main network has its own optimization, where the inlets to subnetworks are modelled as users (see also Optimize allocation per source in a fixed order #565 (comment));
• These 'users' can also be sources and have no return flow. Determining whether there is water left in a subnetwork to give back to the main network is a quite complex problem on its own.

Explicitly introducing this main network closes the door to arbitrarily deeply nested allocation. I would say that this is not a problem because arbitrarily deeply nested allocation can be quite confusing to the modeler in terms of input and behavior, and wont see much use cases.

* I use subnetwork and allocation network somewhat interchangeably. At first there was a clear distinction, with the subnetwork consisting of a subset of nodes, and the allocation network being the 'separate' derived graph used in allocation optimization. However, #732 will remove this distinction. Currently in the in- and output names only allocation network is used.

[SouthEndMusic]

Closed in favor of #920, #921, #922.