fix(iroh): Always update the best addr after changes

[matheus23]

Description

There was a bug where in NodeMap::add_node_addr wouldn't cause an update to the NodeUdpState::best addr, because we wouldn't call NodeUdpState::udpate_best_addr after changes in NodeState::update_from_node_addr.

In an effort to fix this class of bugs once and for all, I've made NodeUdpState::paths private and added guarded access. The access guard has a Drop implementation that calls update_best_addr.

We originally saw this bug as an error message in iroh-blobs tests:

---

While making NodeUdpState's fields private, I noticed that NodeUdpState::best was sometimes accessed instead of NodeUdpState::send_addr. This is now fixed, but required passing through the have_ipv6 boolean quite a bit, so that's why the diff is slightly bigger.
Feel free to look at commits in isolation instead.

Change checklist

• Self-review.
• Documentation updates following the style guide, if relevant.

[github-actions]

Documentation for this PR has been generated and is available at: https://n0-computer.github.io/iroh/pr/3422/docs/iroh/

Last updated: 2025-07-31T14:00:22Z

[github-actions]

Netsim report & logs for this PR have been generated and is available at: LOGS
This report will remain available for 3 days.

Last updated for commit: 062bef3

[flub]

This is cool. Have you considered moving all the logic that needs to have access to the internals here to be methods on the NodeUdpPaths itself and protecting access and ensuring updating of best addr that way? I think it's what I would have gone for, or at least tried. I appreciate this is less changes.

I'm also aware that all these changes interact with the changes needed for multipath... and I haven't wrapped my head around what those are yet so don't know if more or less cleaning up here is helpful.

[matheus23]

Yeah I had considered that. Would've been substantially more changes I guess