lstmp error message enhancement.
sequence_conv error message enhencement.
sequence_enumerate error message enhencement.
sequence_mask error message enhencement.

* test=develop, bug fix

As the title, decrease random failure probability for test_parallel_executor_mnist

The old code set larger delta when comparing reduce and all reduce, but didn't set all. I added it.

On my linux machine, I run 100 times, no failure occurs. In addition, we only saw this random failure on CI two times since I worked. I thought it was rare and I just increased the delta.

* fix error message in conv/conv_transpose. test=develop

* refine err msg in shuffle channel op

* fix approveals message;test=document_fix

* fix approveals message;test=document_fix

* fix tensorrt 6 build error. test=develop

* fix. test=develop

* bug fix

* test=develop

* Remove dependency on nltk for paddle __init__.
test=develop

* Remove nltk.movie_reivew sentiment dataset to remove dependency on nltk.
test=develop

* use iwyu clean include, test=develop, test=win

* compilation error, test=develop

* fix compilation error2, test=develop

* fix compilation error3, test=develop

* fix compilation error4, test=develop

* fix compilation error5, test=develop

* fix compilation error6, test=develop

* fix compilation error7, test=develop

* fix compilation error8, test=develop

* fix compilation error8, test=develop

* fix compilation error10, test=develop

* fix compilation error11, test=develop

* reproduce summary api

* rename StaticLayer

* rename

* refine the error message for bath size like

* fix dtype of gradientclipbyglobalnorm

* fix dtype bug of GradientClipbyGlobalnorm

Add GPU Kernels of Segment Ops,  support, sum, max, min, mean

* init commit of new save/load

* fix failed unittests

* fix save_load_v2 unittest failed

* fix failed unittest & polish doc

* add tests for coverage

* add more tests & move static apis

* fix example code error

* polish emample code

* fix detail example code problem

* [bug fix]:Memory increases after adapting the cudnn version to 8

* [bug fix]cudnnGetConvolutionForwardAlgorithm not defined

* Makes the Bind Method more intelligent

* Makes the Bind Method more intelligent

* fix unittest

* fix unittest

* fix conflict

The calculation result of Diag_v2 Op under large size input is wrong 

* rename communicator meet->BatchesCounter

* fix parame recv for sparse

* geo sparse init from pserver

* optimize init from pserver

* add large scale optimizer fuse(SGD/ADAM)

* rectification init_worker and exe.run startup program

* polish no onwer op error message

* fix unittest failed

* polish details based reviewer comment

* Add example for cache and buffered, test=develop, test=document_fix

* add fused_bn_add_relu op

* polish some lost error msg

* add some math file to white list

* polish detail based reviewer commnet