Thanks for the PR! If you want to get the actual query complexity, I don't think this library is the right tool for the job. The actual query complexity always depends on the returned data which will only ever be available after it has been resolved. So you'd have to add something to the resolver. Maybe I would add a value to the GraphQL context that I'd increment inside of the resolvers (or simplify that process with a middleware).
I am not sure if only the path is really helpful. You can also request fields via aliases that don't match the schema, there can be union / interface types, you can query the same field multiple times with different aliases / different input arguments etc. That information would be gone and could also affect the complexity:

{
  posts1: posts(limit: 100000) {
    title
    relatedPosts(limit: 3) {
      title # Here you would have the path ["posts1", "relatedPosts", "title"] What could you do with this?
    }
  }
  posts2: posts(limit: 1) {
    title
    relatedPosts(limit: 3000000) {
      title
    }
  }
}

Not sure I fully understand the benefit.

Closing this here as there was no activity for a while and I'm not sure this is something we want to add here.