Description

Fixes #3862

The problem was the following on the client side:

• quinn's QNT would establish a hole-punched path at path X, with the status set to backup initially (that's what QNT rightfully does)
• iroh gets the path open event, selects its best path, finds the IP path, and calls RemoteStateActor::open_path
• that function detects that this path is already tracked (?) and skips setting the path status to available
• quinn keeps sending via path 0, because all paths are backup and that's the lowest path ID.

The fix is to actually set the path to available in all cases.

Also in this PR:

• Change debug prints of ?path_id to %path_id to use the display impl.
• Add another logging event: iroh::_event::path::set_status, which I personally found really useful when debugging.
• Updating to quinn main, because otherwise we don't have the required quinn fixes to make hole-punching work correctly, but we need it to work for the regression test I added.
• Adjustments from the quinn update.

Notes & open questions

• Not quite sure why we get into this condition where we have the path tracked, but not set it to backup/available yet.

Change checklist

• Self-review.