Refactor node table in Python

[evetion]

Given node ids are unique again #1717, we can move away from the multinodetable approach, which should simplify our code (and validation) considerably. This would not be breaking, so low priority for now.

[visr]

Right now we have mode.node_table() which creates the Node table on demand, and the actual data is stored per node type under e.g. model.terminal.node.

It would probably indeed simplify things to have the real table in one place. It would probably be breaking to remove model.terminal.node, and getting the node table for one node type is quite convenient as well. So we should see it we can keep that.