fix(trunk): move T8 #[ignore] full-e2e stub to binary crate tests

The #[ignore]'d full_pstn_e2e_through_media_thread_register_trunk test
was in crates/rutster-trunk/tests/sim_integ.rs, but rutster-trunk can't
depend on the rutster binary crate (circular dep: rutster already
depends on rutster-trunk). The test needs MediaThread + spawn_tap_engine
which live in the binary crate. Moved to
crates/rutster/tests/trunk_sim_e2e.rs where it can access those types
when implemented. The active pstn_sim_synthetic_caller_drives_trunk_reflex_loop
test stays in rutster-trunk (covers the FOB reflex loop + trunk-leg tick
directly). Also pinned Cargo.lock deps to 1.85-compatible versions
(icu/time/rcgen/dimpl downgrades).

Signed-off-by: Aaron D. Lee <himself@adlee.work>

crates/rutster-trunk/tests/sim_integ.rs

@@ -330,19 +330,3 @@ async fn pstn_sim_synthetic_caller_drives_trunk_reflex_loop() {
     let _ = close_tx.send(());
     let _ = tokio::time::timeout(Duration::from_secs(1), engine_handle).await;
 }
-
-/// Full end-to-end test using the binary's `MediaThread` + `RegisterTrunk`.
-///
-/// This test is currently ignored because `rutster-trunk` integration tests
-/// cannot depend on the `rutster` binary crate (`rutster` already depends on
-/// `rutster-trunk`; Cargo disallows circular dev-dependencies). The active
-/// `pstn_sim_synthetic_caller_drives_trunk_reflex_loop` above covers the FOB
-// reflex loop + trunk-leg tick directly; this stub marks where the MediaThread
-/// wiring test belongs once the binary crate is ready to exercise it.
-#[ignore]
-#[tokio::test]
-async fn full_pstn_e2e_through_media_thread_register_trunk() {
-    // TODO: exercise MediaCmd::RegisterTrunk + MediaThread tick loop against a
-    // live TwilioMediaStreamsServer mock once the binary crate exposes a test
-    // harness from the appropriate crate.
-}

crates/rutster/tests/trunk_sim_e2e.rs

@@ -0,0 +1,35 @@
+//! Full end-to-end PSTN sim via the binary's `MediaThread` + `RegisterTrunk`.
+//!
+//! This test was moved from `crates/rutster-trunk/tests/sim_integ.rs` because
+//! `rutster-trunk` integration tests cannot depend on the `rutster` binary
+//! crate (`rutster` already depends on `rutster-trunk`; Cargo disallows
+//! circular dev-dependencies). The binary crate has access to
+//! `MediaThread`, `MediaCmd::RegisterTrunk`, and `spawn_tap_engine` — the
+//! full wiring surface the trunk crate's tests couldn't reach.
+//!
+//! The active `pstn_sim_synthetic_caller_drives_trunk_reflex_loop` in
+//! `crates/rutster-trunk/tests/sim_integ.rs` covers the FOB reflex loop +
+//! trunk-leg tick directly. This test extends that to exercise the binary's
+//! `MediaThread` command channel + the `RegisterTrunk` spawn seam end-to-end.
+
+#[ignore = "requires MediaThread::RegisterTrunk wiring + MockTwilioMediaStreamsServer; not yet implemented in the binary crate's test harness"]
+#[tokio::test]
+async fn full_pstn_e2e_through_media_thread_register_trunk() {
+    // TODO: exercise MediaCmd::RegisterTrunk + MediaThread tick loop against a
+    // live TwilioMediaStreamsServer mock. The test should:
+    //   1. Start MockRealtimeBrain.
+    //   2. Spawn a MediaThread (from crates/rutster/src/media_thread.rs).
+    //   3. Construct a MockTwilioMediaStreamsServer that pushes synthetic loud
+    //      PcmFrames into a RegisterTrunkInboundChannel.
+    //   4. Send MediaCmd::RegisterTrunk to the MediaThread.
+    //   5. Drive the 20ms tick loop for ~500ms.
+    //   6. Assert barge-in fired (ReflexMetrics.barge_in_count >= 1).
+    //   7. Assert brain reply observed on the outbound mpsc.
+    //   8. Assert idle timeout closes the session (ChannelState::Closed).
+    //
+    // This requires the binary crate's test harness to expose a way to
+    // construct a TrunkSession's pipe stack outside the normal axum route
+    // handler (or to drive the route handler in-process via axum's test
+    // `oneshot` pattern). Either path is feasible; pick the one that
+    // matches the existing `barge_in_integration.rs` harness pattern.
+}