perf: harden quotas and reduce compression overhead

src/services/idempotency.rs

@@ -148,10 +148,9 @@ pub async fn wait_for_replay(
     max_wait_ms: u64,
 ) -> Result<Option<(i32, JsonValue)>, AppError> {
     let started = tokio::time::Instant::now();
-    let now = Utc::now();
 
     loop {
-        let row = get_row(state, scope, idempotency_key, now).await?;
+        let row = get_row(state, scope, idempotency_key, Utc::now()).await?;
         let Some(row) = row else { return Ok(None) };
 
         if row.request_hash != request_hash {