Fix resolution process

[lquerel]

This PR fixes an issue in the resolution process when there are more than 1 level of extends.

Note 1: The lineage computation is still incomplete and will be fixed in the next PR (see #52). This is not blocking.

Note 2: The current resolution process is iterative and far from optimal but still pretty fast with the current semconv registry. If speed becomes an issue, a graph could be construct from the extends and ref clauses, and a DFS resolution could be used to optimize the entire resolution process.

[jsuereth]

It's interesting how a small fix ripples through all the test code. I think maybe I need to do a better job focusing on the markdown or we need to start pulling in more of the existing markdown as our tests....

IN any case, looks good! Thanks for fixing this!

[jsuereth]

For all these options, you can make a helper method to clean this up.

fn clone_first_some<T : Clone>(first: &Option<T>, second: &Option<T>): Option<T> {
  if first.is_some() {
      first.clone()
  } else {
     second.clone()
  }    

Then all this code becomes:

AttribtueSpec::Ref {
   ...
  brief: clone_first_some(brief, parent_brief),
  examples: clone_first_some(examples, parent_examples),
  ...
}

[lquerel]

Created the following GH issue #54

[lquerel]

It's interesting how a small fix ripples through all the test code. I think maybe I need to do a better job focusing on the markdown or we need to start pulling in more of the existing markdown as our tests....

IN any case, looks good! Thanks for fixing this!

Yes looks like using the existing markdown files is our best option at this point.