Do not fail stage after already done

Ignore failure within stage execution if it already done.
This is to work around the problem which was introduced with
39b04491c05.
With that change we close EventDrivenTaskSource when stage completes,
and as a result of that it may emit an failure event due to internall
processes being cancelled.
Handling of event in `StageExecution.fail()` did close a number of
objects needed for query completion (see createStageExecutionCloser()).
As a result the query could hang.

core/trino-main/src/main/java/io/trino/execution/scheduler/faulttolerant/EventDrivenFaultTolerantQueryScheduler.java

@@ -2415,6 +2415,10 @@ public void abort()
 
         public void fail(Throwable t)
         {
+            if (stage.getState().isDone()) {
+                // stage already done; ignore
+                return;
+            }
             Closer closer = createStageExecutionCloser();
             closer.register(() -> stage.fail(t));
             try {