Support conditions on linear combinations of variables for DiscreteControl

[SouthEndMusic]

Fixes #1346.

Upgrade instructions

This is a breaking change to the DiscreteControl tables, that makes it possible to listen to a weighted combination of multiple variables. Until so far we were only able to listen to one variable, like the water level in a particular Basin.

To listen to the average of four water levels, you can assign a weight of 0.25 to each. To listen to the difference of two water levels, you can assign a weight of 1 to one, and -1 to the other.

For a description of the new table schemas, see https://deltares.github.io/Ribasim/core/usage.html#sec-discrete-control. The combinations of variables are called compound variables, and are identified by a shared compound_variable_id.

[SouthEndMusic]

I Had some problems with the table name CompoundVariable, apparently this is not allowed since the schema name which contains the snake case compound_variable may not contain underscores.

[evetion]

Nice PR! 👍🏻 I wonder if we should change the input tables slightly (but it would be breaking).

Specifically, compoundvariable is now mostly a repeat of condition, with the change of only a single field (added weight, dropped greater_than). Furthermore, condition is changed that type can take "compound", while the node_id is ignored.

I wonder whether we should split condition into condition and variable, where variable can be compound if a id occurs multiple times. listen-*, variable and look ahead are then all dropped/moved.

[SouthEndMusic]

Some clarifications for the latest commits:

• We're now back at allowing multiple compound variables on the same DiscreteControl node using the compound_variable_id. This means that we're back in the situation where all the control of a whole model can be done with a single DiscreteControl node, so it's up to the modeller to do this in a sensible way;
• We also allow multiple greater_than values on a single compound variable, which all define their own condition. I did not want to copy the compound variable data for every condition, so I now store all greater_than values per compound variable in discrete_control.greater_than. This does require some translation between the condition_idx (which is for a single flat dense vector) and the compound variable and greater_than indices, see get_discrete_control_indices.