Improved testing for subnetwork communication

[simulutions]

Continues on #2556 and fixes #2632.

The allocation between subnetworks and the primary network had was developed in #2556 and tested on a minimal model with a single secondary network and 1 route for the flow.

In this PR, a more complex test network was created and showed several bugs which were fixed.
In the model below allocation was calculated once as a single network and also as a primary network connected to two secondary networks.

The inflow from the flow boundary of 0.9 m3/s was exactly matched by the 3 user demands of 0.3 m3/s. The results of allocation were compared with the single network version and matched very well (atol = 1e-8).

[visr]

Great to have more testing.

This also fixes the scaling, right? If you are quite sure about that fix but the new tests are still giving trouble, we can consider merging the fix already before the release (Monday I think).