* update, test=develop

* add multiclass_nms

* add multiclass_nms unittest

* add default enable_tensorrt_oss option

* refine multiclas nms unittest and add serialization/dynamic test

* change super to InferencePassTest for python2 compatibility

* refine multiclass nms unittest

* move out dynamic shape test due to ci timelimit

* delete include framework.pb.h

* fix error

* fix zip py3 bug

* polish two error messages

* polish details

Fix default argument of nn.MaxPool3D()

Our old `loop_body` function may return single element when `loop_vars` just contains only 1 element, which can cause bug. The key point of this PR is forcing `loop_body` functions always return tuple.

* fix cache key in concat oneDNN kernel

* key simplified

* Refine saving output scale to infer program

* modify windows CI to VS2017

* modify windows CI to VS2017

* modify windows CI to VS2017

* add relu forward kernel and backward kernel

* Delete fast_check_nan_inf

* Delete run_fast_nan_inf_debug

* fix tensorrt output varible reshape

* move padding shape x 1 x 1 in ernie to qkv and fc

* update layer name

* fix softmax when input is dynamic, fc not padding any more

* fix varlen

* move fc x_dim assert to op_teller

fix launch ps ut test=develop

* nearest_interp op converter w/ dynamic/static

* fix data_layout include

* add trt nearest unit_test

* add nearest_interp NHWC test

* update trt nearest interp nhwc testcase

* remove asterisk for python2 compatibility

* add empty line to prevent conflict

* nearest_interp op converter w/ dynamic/static

* fix data_layout include

* add trt nearest unit_test

* add nearest_interp NHWC test

* update trt nearest interp nhwc testcase

* remove asterisk for python2 compatibility

* add empty line to prevent conflict

* change the priority of out_h, out_w

* add 1f1b scheduler for pp, test=develop

* Fix skip_quant in QAT

* support float16 in custom op

* fix failed unittests

* support attribute in infershape

* polish details