Read NetCDF files with demand_priority

[visr]

This extends NetCDF reading support (#2434), specifically the demand_priority dimension, so these tables:

• UserDemand / time (demand_priority, node_id, time)
• LevelDemand / time (demand_priority, node_id, time)
• FlowDemand / time (demand_priority, node_id, time)

This PR contains some more refactoring of our NetCDF reading code to be able to handle files written by Delft-FEWS. This means e.g. recognizing coordinates by their cf_role or standard_name first, rather than only the variable name. For instance FEWS wouldn't write node_id, but stations with cf_role = timeseries_id.

This still needs more testing, but since I cannot do that now, I made #2617 and propose we add these later.

[evetion]

I feel that 99% of the code here could be upstreamed to a table interface implementation on NCDatasets. Without significant tests, I'm hesitant to merge this, especially given the inner/outer repeats that are prone to bugs.

[evetion]

👀