Fix abort issue in cpu multi-threads

[tensor-tang]

fix #11231

The key backtrace is :

#16 0x00000000029b4640 in paddle::framework::Scope::Var (this=0x5cfd0e0, name="lstm_0.w_0")
    at /home/tangjian/paddle-docker/paddle/fluid/framework/scope.cc:55
#17 0x0000000002705d91 in paddle::framework::Executor::CreateVariables (this=0x7f07caffcab0, pdesc=...,
    scope=0x7f07b8022850, block_id=0) at /home/tangjian/paddle-docker/paddle/fluid/framework/executor.cc:94
#18 0x0000000002709942 in paddle::framework::Executor::RunPreparedContext (this=0x7f07caffcab0, ctx=0x7f07b8022990,
    scope=0x5cfd0e0, create_local_scope=true, create_vars=true)
    at /home/tangjian/paddle-docker/paddle/fluid/framework/executor.cc:316
#19 0x0000000002706597 in paddle::framework::Executor::Run (this=0x7f07caffcab0, pdesc=..., scope=0x5cfd0e0,
    block_id=0, create_local_scope=true, create_vars=true)
    at /home/tangjian/paddle-docker/paddle/fluid/framework/executor.cc:119
#20 0x00000000022e875e in paddle::inference::LoadPersistables (executor=0x7f07caffcab0, scope=0x5cfd0e0,
    main_program=..., dirname="/home/tangjian/nlp/epoch6", param_filename="")
    at /home/tangjian/paddle-docker/paddle/fluid/inference/io.cc:108
#21 0x00000000022e8c34 in paddle::inference::Load (executor=0x7f07caffcab0, scope=0x5cfd0e0,
    dirname="/home/tangjian/nlp/epoch6") at /home/tangjian/paddle-docker/paddle/fluid/inference/io.cc:124
#22 0x0000000002244eed in ThreadRunInfer (tid=5, scope=0x5cfd0e0, jobs=std::vector of length 6, capacity 6 = {...})
    at /home/tangjian/paddle-docker/paddle/fluid/inference/tests/book/test_inference_nlp.cc:111

[tensor-tang]

https://paddleci.ngrok.io/viewLog.html?tab=buildLog&logTab=tree&filter=debug&expand=all&buildId=38034&_focus=5958

test_listen_and_serv_op .........................***Timeout 31.59 sec

[panyx0718]

paddle::platform::CPUPlace place;
paddle::framework::Executor executor(place);

[panyx0718]

same as above