fix(relay): format the new import, and bound the mint wait

Two review findings from Greptile on the contributed change, neither about the
fix itself:

- `cargo fmt --all --check` failed: `use tokio::sync::Notify` sat inside the
  futures import group, before the `nostream_config` block, so the required
  format step would have stopped the checks before the test ran.
- `create_started().await` had no deadline, so a handler that panicked before
  minting would hang the test instead of failing it. The wait is now wrapped in
  a five-second timeout with a message that says which wait expired.

`cargo fmt --all --check` is clean, and
`cargo test -p nostream-relay --test handlers a_close_during_the_invite_mint`
passes.

Author: chappie-daemon

crates/nostream-relay/src/testing.rs

@@ -16,7 +16,6 @@ use std::sync::{Arc, Mutex};
 
 use futures_util::StreamExt;
 use futures_util::stream;
-use tokio::sync::Notify;
 use nostream_config::settings::{
     Info, Network, Nip42Settings, Nip43Settings, Nip45Settings, Nip50Settings, Nip56Settings,
     RateLimiterSettings, Settings,
@@ -30,6 +29,7 @@ use nostream_storage::repositories::{
     NewReport, Nip05Verification, Nip05VerificationRepository, QueryStream, ReconciliationItem,
     Report, ReportRepository, StorageFuture, User, UserPatch, UserRepository,
 };
+use tokio::sync::Notify;
 
 use crate::connection::{Connection, Subscription};
 use crate::error::RelayError;

crates/nostream-relay/tests/handlers.rs

@@ -1263,7 +1263,9 @@ async fn a_close_during_the_invite_mint_leaves_the_event_but_not_the_read() {
     );
 
     // The mint's one `await` is held, which is the window a `CLOSE` lands in.
-    fixture.invites.create_started().await;
+    tokio::time::timeout(Duration::from_secs(5), fixture.invites.create_started())
+        .await
+        .expect("the invite mint starts within five seconds");
 
     assert_eq!(
         fixture.invites.creates_started(),