[BUG] Three winding transformers are represented as cycles in the graph

[vincentkoppen]

Describe the bug
The support within the graphs for three windings arrays has some strange side effect currently for some of the graphs functions.

Below shows the identified problems

   grid = Grid.empty()

    node_array = NodeArray.empty(3)
    node_array.id = [1, 2, 3]
    grid.append(node_array)

    three_winding_array = ThreeWindingTransformerArray.empty(1)
    three_winding_array.node_1 = 1
    three_winding_array.node_2 = 2
    three_winding_array.node_3 = 3
    three_winding_array.status_1 = 1
    three_winding_array.status_2 = 1
    three_winding_array.status_3 = 1
    grid.append(three_winding_array)

    cycles = grid.graphs.active_graph.find_fundamental_cycles()
    # cycles = [[2, 1, 3, 2]]
    paths = grid.graphs.active_graph.get_all_paths(1, 2)
    # paths = [[1,3,2], [1,2]]

[vincentkoppen]

From a quick think, there is no easy fix possible at this point. Sometimes you want to represent a three winding as a cycle, sometimes as a star (and then ignore the dummy node that we have to introduce for it).
Discussed with @nitbharambe and so far we might need two graph representations internally (one where it is a cycle, one where it is a star) and use the correct one for each algorithm.

This change would include:

• Introduce a second graph implementation
• Update all add/delete functions to match this graph implementation
• Add support for a dummy point (that does not have an extenal_id)
• Use the new graph representation within get_all_paths and find_fundamental_cycles
  • Make sure to not include the dummy points

[vincentkoppen]

There is no easy solution as far as we can see.
There is a work in progress setup currently that should fix all the issues. Main work that is left for this is:

• Cleanup of the code
• Adding tests for all the different situations