[fix](cloud-mow) duplicate key occured when StreamLoad commit failed …

…and retry (apache#36670)

## Proposed changes

Issue Number: close #xxx

In our stress test environment, we found that for cloud mow table, when
stream load commit failed and retry, the StreamLoadContext->commit_infos
is cleared, and FE skipped the delete bitmap calculation task and commit
success directly, in such case, the delete bitmap is not calculated
correctly

be/src/runtime/stream_load/stream_load_executor.cpp

@@ -294,7 +294,7 @@ void StreamLoadExecutor::get_commit_request(StreamLoadContext* ctx,
     request.tbl = ctx->table;
     request.txnId = ctx->txn_id;
     request.sync = true;
-    request.commitInfos = std::move(ctx->commit_infos);
+    request.commitInfos = ctx->commit_infos;
     request.__isset.commitInfos = true;
     request.__set_thrift_rpc_timeout_ms(config::txn_commit_rpc_timeout_ms);
     request.tbls = ctx->table_list;