test: fix flaky TestStatusHandler_RecentRepos_FailureHasErrorFields (#24)

The test waited on `len(RecentRepos) > 0`, but a repo is tracked (and thus
present in RecentRepos) as soon as processing starts — before the handler
returns and the queue records last_error/last_error_time. Under load the status
snapshot could be read in that window, so `last_error` came back empty.

Wait on the actual recorded signal instead: `RecentRepos[0].LastError != ""`
for the failure test, and `!RecentRepos[0].LastSuccessTime.IsZero()` for the
sibling success test (TestStatusHandler_ReportsPerRepoRecentRepos), which had
the same latent race against `last_success_time`.

Reproduced reliably with `-race`; fixed test now passes 40x under `-race`.
Production sync-queue code is unchanged.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

Author: tae2089

cmd/ccg/health_test.go

@@ -373,8 +373,11 @@ func TestStatusHandler_ReportsPerRepoRecentRepos(t *testing.T) {
 		t.Fatal("timed out waiting for handler")
 	}
 
+	// The repo appears in RecentRepos as soon as it is tracked, before the handler
+	// returns and the success time is recorded. Wait for the recorded success.
 	for deadline := time.Now().Add(2 * time.Second); time.Now().Before(deadline); {
-		if len(q.Stats().RecentRepos) > 0 {
+		rr := q.Stats().RecentRepos
+		if len(rr) > 0 && !rr[0].LastSuccessTime.IsZero() {
 			break
 		}
 		time.Sleep(10 * time.Millisecond)
@@ -434,8 +437,11 @@ func TestStatusHandler_RecentRepos_FailureHasErrorFields(t *testing.T) {
 		t.Fatal("timed out waiting for handler")
 	}
 
+	// The repo appears in RecentRepos as soon as it is tracked, before the handler
+	// returns and the failure (last_error) is recorded. Wait for the recorded error.
 	for deadline := time.Now().Add(2 * time.Second); time.Now().Before(deadline); {
-		if len(q.Stats().RecentRepos) > 0 {
+		rr := q.Stats().RecentRepos
+		if len(rr) > 0 && rr[0].LastError != "" {
 			break
 		}
 		time.Sleep(10 * time.Millisecond)