fix: relative-distance corrections

[vaeng]

          I've been trying to implement this one in Elixir, but I'm having trouble solving it, I think some cases may have issues.

1. "Direct parent-child relation" => degree 1, no problem
2. "Sibling relationship" => expected degree between siblings is 1, but from the description I expected it to be 2, with the path being: left child => parent => right child. Bypassing the parent for siblings is fine as a rule, but it should be explicit.
3. "Two degrees of separation, grandchild" => degree 2, no problem
4. "Unrelated individuals" => no problem
5. "Complex graph, cousins" => degree 4, but my solution gave degree 9. I did it by hand and I also get 9 (with the rule of siblings having a degree of separation 1, otherwise you would get 10)
6. "Complex graph, no shortcut, far removed nephew" => degree 15, but my solution fives me 14 (it could be 15 without the siblings rule potentially, I did not do this one by hand)
7. "Complex graph, some shortcuts, cross-down and cross-up, cousins three times removed" => degree 8, but my solution gives me no connection, and I confirmed by hand. Sofia doesn't have a parent, and she starts a second family with Qi in it.

@vaeng could you help me here?

Also in general, I like having large inputs, but only at the end when all base case have been covered. The cousin case, for example could be much smaller, it took me a long time to verify by hand.
I also noticed that the Elixir exercise generator does not preserve the ordering of the family tree entries, but with random order I never would have been able to check my solutions by hand, so maybe it's worth adding a comment saying that preserving the order would be a good idea.

Here is my tentative implementation.

Originally posted by @jiegillet in #2537 (comment)

[vaeng]

Thank you for posting @jiegillet. I moved this to a new issue.

I was implementing this for C++ and had similar thoughts. I have to admit that I did not solve my solution programmatically and analyzed a mermaid graph I made by hand.

We should fix the tests (with new uuids etc).

[ErikSchierboom]

Don't use new uuids. It is very unlikely that maintainers have already implemented the exercise and it'll just be messy. Ping me to force merge the fix PR

[ErikSchierboom]

Any update on this @vaeng ? Would be a cool exercise to add to several tracks

[vaeng]

@jiegillet

My C++ implementation has the same results as yours. I will suggest edits for the spec today and tag you.

[vaeng]

I'd like to merge these before other people try and fail. @jiegillet @ErikSchierboom can any of you verify?

[jiegillet]

I have already verified on the PR, haven't it? Data is right, there is a typo in one exercise name, and I suggested to make the sibling rule explicit in the instructions.

[ErikSchierboom]

@vaeng Maybe update the suggestions from jiegillet? And then we can probably merge