test: run node-level mesh tests over an in-process loopback transport

Add a Loopback variant to TransportHandle backed by an unbounded
in-process channel and a shared address-to-receiver registry, so
node-level multi-node tests deliver packets directly between nodes
instead of over real localhost UDP sockets. This removes the kernel
UDP receive-buffer overflow that dropped handshake packets when many
tests ran in parallel under CPU contention, and lets the large-network
convergence tests run reliably in the default suite again (their
parallel-load ignore markers are removed).

The new transport and its enum variant are cfg(test)-gated, so the
daemon build is unaffected.

src/node/tests/session.rs

@@ -571,7 +571,6 @@ async fn drain_to_quiescence(nodes: &mut [TestNode]) {
 }
 
 #[tokio::test]
-#[ignore = "parallel-load flake class — re-enable when fixed (run solo with --ignored or --test-threads=1 in the meantime)"]
 async fn test_session_100_nodes() {
     let _guard = lock_large_network_test().await;
 
@@ -1252,7 +1251,6 @@ async fn test_tun_outbound_3node_forwarded() {
 }
 
 #[tokio::test]
-#[ignore = "parallel-load flake class — re-enable when fixed (run solo with --ignored or --test-threads=1 in the meantime)"]
 async fn test_tun_outbound_pending_queue_flush() {
     // Send multiple packets before session exists — all should be delivered
     let edges = vec![(0, 1)];