fix: propagate runAsync dispatch failure into retryFuture

If ForkJoinPool.commonPool() rejects the retry task (e.g. during JVM
shutdown), the runAsync future completes exceptionally but retryFuture
was never completed, causing the overall request future to hang. Wire
an .exceptionally() handler on the runAsync future so any dispatch
failure propagates into retryFuture immediately.

Author: allanrogerr

api/src/main/java/io/minio/BaseS3Client.java

@@ -321,13 +321,18 @@ private CompletableFuture<Response> executeWithRetry(
               RETRY_SCHEDULER.schedule(
                   () ->
                       CompletableFuture.runAsync(
-                          () ->
-                              executeWithRetry(s3request, region, maxAttempts, attempt + 1)
-                                  .whenComplete(
-                                      (r, t) -> {
-                                        if (t != null) retryFuture.completeExceptionally(t);
-                                        else retryFuture.complete(r);
-                                      })),
+                              () ->
+                                  executeWithRetry(s3request, region, maxAttempts, attempt + 1)
+                                      .whenComplete(
+                                          (r, t) -> {
+                                            if (t != null) retryFuture.completeExceptionally(t);
+                                            else retryFuture.complete(r);
+                                          }))
+                          .exceptionally(
+                              ex -> {
+                                retryFuture.completeExceptionally(ex);
+                                return null;
+                              }),
                   delayMs,
                   TimeUnit.MILLISECONDS);
               return retryFuture;