Add a separate method for rerouting with reset failed counter (#115896)

A cluster resets failed counter and calls reroute on node-join. This is
a background activity not directly initiated by end-users. Currently it
uses the same reroute method that handles the reoute API. This creates
some ambiguity on the method's scope. This PR adds a new dedicate method
for this use case and leave the existing one dedicated for API usage.

Author: ywangd

server/src/main/java/org/elasticsearch/cluster/routing/allocation/AllocationService.java

@@ -572,7 +572,7 @@ public void addAllocFailuresResetListenerTo(ClusterService clusterService) {
         // set retryFailed=true to trigger failures reset during reroute
         var taskQueue = clusterService.createTaskQueue("reset-allocation-failures", Priority.NORMAL, (batchCtx) -> {
             batchCtx.taskContexts().forEach((taskCtx) -> taskCtx.success(() -> {}));
-            return reroute(batchCtx.initialState(), new AllocationCommands(), false, true, false, ActionListener.noop()).clusterState();
+            return rerouteWithResetFailedCounter(batchCtx.initialState());
         });
 
         clusterService.addListener((changeEvent) -> {
@@ -582,6 +582,13 @@ public void addAllocFailuresResetListenerTo(ClusterService clusterService) {
         });
     }
 
+    private ClusterState rerouteWithResetFailedCounter(ClusterState clusterState) {
+        RoutingAllocation allocation = createRoutingAllocation(clusterState, currentNanoTime());
+        allocation.routingNodes().resetFailedCounter(allocation.changes());
+        reroute(allocation, routingAllocation -> shardsAllocator.allocate(routingAllocation, ActionListener.noop()));
+        return buildResultAndLogHealthChange(clusterState, allocation, "reroute with reset failed counter");
+    }
+
     private static void disassociateDeadNodes(RoutingAllocation allocation) {
         for (Iterator<RoutingNode> it = allocation.routingNodes().mutableIterator(); it.hasNext();) {
             RoutingNode node = it.next();