scaleset: reap timed-out runners stuck in failed status

reapTimedOutRunners only considered runners in pending/installing
RunnerStatus as eligible for reaping. A runner whose agent reports a
failure (RunnerStatus set to failed via AddInstanceStatusMessage) was
never pending/installing again, and would never register as idle/active
in github either, so it was permanently excluded from this check and
leaked forever (instance never transitioned to pending_delete).

Allow RunnerFailed through the same eligibility check. Downstream
logic still requires the runner be missing from github or reported
offline there before reaping, unchanged.

Author: benoit-nexthop

workers/scaleset/scaleset.go

@@ -395,8 +395,8 @@ func (w *Worker) reapTimedOutRunners(runners map[string]params.RunnerReference)
 			continue
 		}
 
-		if runner.RunnerStatus != params.RunnerPending && runner.RunnerStatus != params.RunnerInstalling {
-			slog.DebugContext(w.ctx, "runner is not pending or installing; skipping", "runner_name", runner.Name)
+		if runner.RunnerStatus != params.RunnerPending && runner.RunnerStatus != params.RunnerInstalling && runner.RunnerStatus != params.RunnerFailed {
+			slog.DebugContext(w.ctx, "runner is not pending, installing or failed; skipping", "runner_name", runner.Name)
 			continue
 		}
 		if ghRunner, ok := runners[runner.Name]; !ok || ghRunner.GetStatus() == params.RunnerOffline {