fix: terminate ephemeral and location loop immediately on channel close

When scheduler is destroyed, e.g. during a key import,
there is some time between destroying the interrupt channel
and the loop task.

To avoid busy looping, tasks should terminate if
receiving from the interrupt loop fails
instead of treating it as the interrupt.

src/ephemeral.rs

@@ -569,9 +569,21 @@ pub(crate) async fn ephemeral_loop(context: &Context, interrupt_receiver: Receiv
                 "Ephemeral loop waiting for deletion in {} or interrupt",
                 duration_to_str(duration)
             );
-            if timeout(duration, interrupt_receiver.recv()).await.is_ok() {
-                // received an interruption signal, recompute waiting time (if any)
-                continue;
+            match timeout(duration, interrupt_receiver.recv()).await {
+                Ok(Ok(())) => {
+                    // received an interruption signal, recompute waiting time (if any)
+                    continue;
+                }
+                Ok(Err(err)) => {
+                    warn!(
+                        context,
+                        "Interrupt channel closed, ephemeral loop exits now: {err:#}."
+                    );
+                    return;
+                }
+                Err(_err) => {
+                    // Timeout.
+                }
             }
         }