Fix num_threads settings in fast_pe

Revert "Softmax op optimization for inference "

add self to author

test=develop

Softmax op optimization for inference 

test=develop

test=develop

test=develop

fix the nn.py example

* Support softmax return in softmax_with_cross_entropy

* Add test for return_softmax=False

test=develop

test=develop

speedup DetectPatterns

Density prior box operator for image detection model.

test=develop

add bilinear_tensor_product layer

test=develop

Set serial for tensorrt utest and set unused logs invisible.

* Add int type support for stack_op

* Improve gather op to support index with shape N x 1

test=develop

* Fix stack_op kernel's registry

test=develop

Adding Intel nGraph build

test=develop

test=develop

test=develop

dist table support other optimize and regular config

reduce pass num of test_label_semantic_roles to avoid test timeout

test=develop

test=develop

test=develop

test=develop

test=develop

test=develop

test=develop

skip mkldnn related pass when use_mkldnn=false to speedup analysis

Add similarity focus op