Fix duplicate boundaries in Delwaq generation.

[evetion]

Fixes #1901
Fixes #1902

This disables any dispersion in the Delwaq model, fixing some tracer intrusions.

#1902 was actually two issues:

• In the graph simplification, we didn't anticipate boundary to boundary connections (UserDemand to Terminal, in use at IJmuiden). We now just remove the Terminal node/edge from it.
• UserDemands are the only boundaries that occur twice (when not connecting back to the same Basin, those are merged to one edge), and that results in a duplicate boundary name, as it was based on type + node_id.

[visr]

Looks good to me overall. I think we should remove the 3 print statements, since this is a valid Ribasim model.
Is there a way to test this easily? Should we for instance read the log you linked in #1902 in test_delwaq.py and ensure that WARNING does not occur or something like that?

[evetion]

I will make the prints debug, and the warnings errors, that should trigger one such a thing happens again in the tests.

[evetion]

Should we for instance read the log you linked in #1902 in test_delwaq.py and ensure that WARNING does not occur or something like that?

There are several warnings (that we can safely ignore) in the output file, so that's hard to test. I'd say if it runs, it runs 🤷🏻 (and we can parse the output). Let's revisit this when we bundle/wrap Delwaq.