[cherry-pick]fix bug in pool/conv/conv_transpose: UpdatePaddingAndDilation, _get_padding_with_SAME and conv2dtranspose_forward_naive. (#20997) (#21225)

* fix bug in pool/conv/conv_transpose:
    1. It should be stride[i] not stride[0] in UpdatePaddingAndDilation;
    2. fix bug of func  _get_padding_with_SAME in test_conv/conv_transpose_op.py;
    3. fix bug of the computation process in function conv2dtranspose_forward_naive.
    test=release/1.6

test=release/1.6

TODO: fix cudnn_conv and re-enable it

test=develop
test=release/1.6

* Add asymetric padding support for mkldnn pooling
test=develop

* Add asymetric padding support for mkldnn conv
test=develop

* Add asymetric padding support for mkldnn conv_transpose
test=develop

[cherry-pick] fix squared_mat_sub_fuse_pass bug when elementwise_op input is persistable param test=develop test=release/1.6 (#21044)

fix squared_mat_sub_fuse_pass bug when elementwise_op input is persistable param

* fix uniform random; test=develop

* add uniform random test; test=develop

test=release/1.6

* fix infer crashes caused by conv/pool upgrades, test=release/1.6

* fix bug, test=release/1.6

fix the bug of conv_transpose cudnn kernel：cherry-pick #20958

Paddle-trt inference: filter conv, depthwise_conv, pooling when padding size > 4
fix C++ multicard  inference bug.
test=develop

* [cherry-pick]fix bug in reshape: (#20781)

consider the situation that shape of input can contain more than one -1.

* [cherry-pick]support Tensor for split and concat, support -1 in num_or_sections, add check num_or_sections (#20780)

* improve split and concat op:
1. support Tensor for argument 'dim' in split op.
2. support Tensor for argument 'axis' in concat op.
* redefine function GetDataFromTensor and set unknown output shape to - 1.
* add check: Attr(sections) match Input(X).
* support Tensor for attr(sections) and attr(sections) can contain -1.
* modify error message and fix bug for concat and call Resize only when necessary.
test=release/1.6

* [cherry-pick]improve unsqueeze op to support int, Tensor for argument axes (#20824)

* improve unsqueeze op to support int, Tensor and Tensor list for argument axes.
* call Resize only when necessary. test=release/1.6

* [cherry-pick]Compatible int32 and int64 for attr in concat/split/unsqueeze. test=release/1.6 (#20912)

cherry-pick1.6 simplify master+patch，remove ins when size != merge_size or has conflict slot  (#20941)

* simplify master+patch，remove ins when size != merge_size or has conflict slot
* test=develop

* add check nan / inf in downpour worker during training
* test=develop

* update pserver decay blocks

* update distributed notify handler

[cherry-pick] keep the size of symmetric padding is 2 for 2d and 3 for 3d. test=release/1.6 (#20903) (#20939)

* Fix Paddle Cloud role maker (#20860)

- fix jit::matmul bug 

input x, shape(m, k), weight, shape(k, n) 

* bug fix, test=develop

* Simplify third_party path in Python package
* Fix typo

[cherry-pick] fix the bug of conv_transpose: compitable with AnyLayout setting, test=release/1.6 #(20897) (#20918)

Bug Fix: Paddle-TRT cannot handle adaptive pooling in pool2d op converter and "num" attribute in split op converter (#20733) (#20902)

* fix pool2d trt converter, test=develop

* add fix for split op converter, test=develop

* add support to gcc8, add docker env
* remove the warning issue

* Serialize to pickle format (#20820)

test=develop

* save load problem fix and new feature add (#20823)

* fix persistable;

* fix save load bugs; test=develop

* fix bug; test=develop

* add example for new io api; test=develop

* addd example; test=develop

* Add IndicateVarDataType interface to block tensor is not initialized problem in OP GetExceptedKernelType (#20044)

* add indicate_var_data_type inferface, test=develop

* add unittests & polish error message, test=develop

* remove needless include, test=develop

* extract public function & polish message, test=develop

* delete empty var check, test=develop

* change data_type to pointer parameter, test=develop

* polish details, test=develop

* Replace risky GetInputType method with secure IndicateVarDataType interface (#20668)

* replace part of the old implementation, test=develop

* restore concat op, test=develop

* update all ops implemention & delete GetDataTypeOfVar func, test=develop

test=release/1.6

* test=document_fix

* test=release/1.6

* add more err msg, test=develop

* add more unittests, test=release/1.6

test=release/1.6 test=document_fix