It seems that the blocking queue might be destroyed early than Run
method complete. It might because the Run method throw some unhandled
exception. However, it should be shared_ptr when multthread access an
resource. So change BlockingQueue as a shared_ptr.

test=develop

fix sequence_pad example error

* fix use_reader_alloc uts

* dist ut fixes test=develop

* update test=develop

* fix test for py3 test=develop

test=develop

remove the optimized but buggy implementation

test=develop

Fix dynamic_gru h_0 bug

Fix Mac Python3 CI job

Max Sequence pool optimization

remove with_inference option

test=develop

test=develop

* feat(platform): lazy initialization of devicecontext in pool

Use std::async(deferer, []{...}) to lazy initialize DeviceContext in Pool

test=develop

* Add future includes

test=develop

test=develop

test=develop

increase test timeout coverage.

1.add position encoding  2.logloss in nn.py

test=develop

test=develop

test=develop

Fix inference model not found on Mac CI

* fix sum_op's GetExpectedKernelType
test=develop

* fix ci fail
test=develop

test=develop

test=develop

test=develop

generate proposal labels doc

Refine/jit/crf decoding

Fix import plot in py3

disable some tests

[1.1] fix cmake warning when ON_INFER=false

OrderSet default is mutableSet, not set.

fix xxhash compile on macos

[1.1] Cpu for 1.1 merge with shape