fix(116393): 递归扁平化 workflow.nodes 修复嵌套子节点排序问题

之前的修复仅使用 workflow.getNodes() 顶层索引，导致并行/包容/人工分支内的
嵌套子节点（如审批节点）被推到结束节点之后。

修复：新增 buildNodeIndexMap() 递归遍历 workflow.nodes 及各 IBlockNode 的 blocks，
为所有节点（含嵌套子节点）建立完整的索引映射，使分支内部审批节点能正确排列在
父分支控制节点与结束节点之间。

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Author: claude

flow-engine-framework/src/main/java/com/codingapi/flow/service/impl/FlowProcessNodeService.java

@@ -242,16 +242,27 @@ public List<ProcessNode> loadNextNode(FlowSession flowSession) {
         }
 
         private List<ProcessNode> sortByWorkflowNodeOrder(List<ProcessNode> processNodes) {
-            List<IFlowNode> workflowNodes = workflow.getNodes();
             Map<String, Integer> indexMap = new HashMap<>();
-            for (int i = 0; i < workflowNodes.size(); i++) {
-                indexMap.put(workflowNodes.get(i).getId(), i);
-            }
+            int[] counter = {0};
+            buildNodeIndexMap(workflow.getNodes(), indexMap, counter);
             return processNodes.stream()
                     .sorted(Comparator.comparingInt(
                             n -> indexMap.getOrDefault(n.getNodeId(), Integer.MAX_VALUE)))
                     .toList();
         }
+
+        private void buildNodeIndexMap(List<IFlowNode> nodes, Map<String, Integer> indexMap, int[] counter) {
+            if (nodes == null || nodes.isEmpty()) {
+                return;
+            }
+            for (IFlowNode node : nodes) {
+                indexMap.put(node.getId(), counter[0]++);
+                List<IFlowNode> blocks = node.blocks();
+                if (blocks != null && !blocks.isEmpty()) {
+                    buildNodeIndexMap(blocks, indexMap, counter);
+                }
+            }
+        }
     }