connectivity refactor

[SouthEndMusic]

Fixes #814.

[SouthEndMusic]

@visr regarding #825: I had 2 tests sometimes failing which I think I now fixed:

• One from expand logic mapping. I saw that the behavior there depended on the order of a loop over a dictionary, I fixed that.
• One from allocation. I tested the exact flow values and allocated values coming out of the LP solve. The allocated values are always the same, the flows not. I suspect that that is because the exact value of some flows does not matter for the minimization of the objective function. Still interesting that the solver behaves non-deterministically. I decided to stop testing the exact flow values.

[Hofer-Julian]

I had a few comments.
Apart from them, it looks good to me.

[Hofer-Julian]

Suggested change

	for i in 1:length(flow_vertical_dict_inverse)
	id = flow_vertical_dict_inverse[i]
	push!(from_node_id, id.value)
	push!(to_node_id, id.value)
	push!(unique_edge_ids_flow, missing)
	end
	for id in flow_vertical_dict_inverse
	push!(from_node_id, id.value)
	push!(to_node_id, id.value)
	push!(unique_edge_ids_flow, missing)
	end

[Hofer-Julian]

Suggested change

	for i in 1:length(flow_dict_inverse)
	from_id, to_id = flow_dict_inverse[i]
	push!(from_node_id, from_id.value)
	push!(to_node_id, to_id.value)
	push!(unique_edge_ids_flow, graph[from_id, to_id].id)
	end
	for (from_id, to_id) in flow_dict_inverse
	push!(from_node_id, from_id.value)
	push!(to_node_id, to_id.value)
	push!(unique_edge_ids_flow, graph[from_id, to_id].id)
	end

[SouthEndMusic]

Regarding this and previous one: I deliberately do not do this, to ensure the order is correct

[Hofer-Julian]

So you verified that iterating through it and indexing one-by-one doesn't behave the same?