Release lock on CoordinatorAddedAsWorkerNode()

Because master_add_node(or others) might acquire ExclusiveLock
and their initiated sessions may call CoordinatorAddedAsWorkerNode().

With this we prevent potential deadlocks.

src/backend/distributed/commands/table.c

@@ -413,8 +413,8 @@ PreprocessAlterTableStmt(Node *node, const char *alterTableCommand,
 	}
 
 	if (processUtilityContext != PROCESS_UTILITY_SUBCOMMAND &&
-		AlterTableDefinesFKeyBetweenPostgresAndNonDistTable(alterTableStatement) &&
+		CoordinatorAddedAsWorkerNode() &&
-		CoordinatorAddedAsWorkerNode())
+		AlterTableDefinesFKeyBetweenPostgresAndNonDistTable(alterTableStatement))
 	{
 		/*
 		 * We don't process subcommands generated by postgres.

src/backend/distributed/operations/worker_node_manager.c

@@ -418,17 +418,13 @@ NodeIsPrimaryWorker(WorkerNode *node)
 
 /*
  * CoordinatorAddedAsWorkerNode returns true if coordinator is added to the
- * pg_dist_node. This function also acquires ShareLock on pg_dist_node
+ * pg_dist_node.
- * and does not release it to ensure that existency of the coordinator in
- * metadata won't be changed until the end of transaction.
  */
 bool
 CoordinatorAddedAsWorkerNode()
 {
 	bool groupContainsNodes = false;
 
-	LockRelationOid(DistNodeRelationId(), ShareLock);
-
 	PrimaryNodeForGroup(COORDINATOR_GROUP_ID, &groupContainsNodes);
 
 	return groupContainsNodes;