(1) set fleet_send_batch_num a default value according to trainer num， the previous 80000 is fixed，if trainer num is much less or larger than 100，global shuffle may have timeout error.

(2) fix load one table bug, add barrier

GPU allocation uses fraction of available memory, also fix the GetUsed without lock

* fix security issue, test=develop

* bug fix, test=develop

* throw an exception when null pointer data with non-zero length PaddleBuf is passed, test=develop

* Fix Mask rcnn predictor
    1. refine memory optim algorithm to support the model with the block op.
    2. output diff : modify the affine channel fuse
    3. add condition_block_infer op
add interface for setting trt calib table dir
test=develop

* add the missing files.
test=develop

* 1 add trt fp16 support
test=develop

* update parallel.py
test=develop

* support center loss
* change tensor copy  api to high level api tensorcopy

* test=develop rewrite the center_loss cuda_kernel to make it faster
and add document of the center loss api,also update test function

* test=document_preview test=develop
update document of center loss

* test=document_preview test=develop
modify API.spec modify test code remove nouse const_cast

Update conv2d transpose link

test=develop

Optimize the error report information when loadcombine fail to open model files test=develop (#18888)

* upgrade ngraph to v0.24 test=develop

* simplify io test=develop

make dist unit test exclusive run

test=develop

test=develop

- optimization of TID to string

test=develop

Revert "use static variable to do cache instead of thread local in thread frequent switching case (#18428)" (#18879)

This reverts commit ce38bb53.

test=develop

test=develop

* remove legacy memory optimization codes, test=develop

* follow huihuang's comments,test=develop

* follow luotao's comments, test=develop

* dump slot

* test

* proto

* dump slot

* test

* proto

* code style

* code style

* code style

* style

* add delete after unseen days

* add unseen days

* code style

* conflict solve
test=develop

* add clear model

* code style
test=develop

* code style
test=develop

* replace link

* update api.spec

* fix mistake

Also fix a dependency error which may cause compile error

* open fuse optimization ops
test=develop

test=develop

test=develop

* first version memory optimize pass, test=develop

* remove move_tensor_sharing_pass, test=develop

* refine code comments, add unittests, test=develop

* turn off memory_optimize by default, test=develop

* follow huihuang's comments, test=develop

* follow chengduoZH's comments, test=develop

* fix grammar error, add const qualifier, fix pass_test exception message, test=develop

* follow chengduoZH's comments 2nd, test=develop

* test=develop fix cpu roi_align_op backward bug

* fix logical APIs

test=develop

test=document_preview

* fix isfinite

* update matmul comments

* update API.spec

test=document_preview

test=develop

* update API.spec

test=document_preview

test=develop

* update API.spec

test=document_preview

test=develop

refine launch_ps and role_maker

test=develop

Fix FleetWrapper:
1. fix shrink dense: just scale show
2. add datanorm scale: divide datanorm's gradient by batch_size

* change INT8 to template so that checking dst_dt with if-else could be removed. CI will be enabled after fixing reviews

* reverse user_residual_memory_p and user_bias_memory_p declaration scope
test=develop

* split test_dist_se_resnext.py into 4 testcases

* extend matmul op to support multiple head multiplication

With the support of multiple head, the multiplication of two big matrixes is
split into multiplication of several (head_number) small matrixes. e.g. if
Mat A is [3, 24] and Mat B is [24, 4], when multiple A and B with head_number
as 4, Mat A will be split as 4 matrix of [3, 6] and Mat B will be 4 matrix of
[6, 4]. The result of final matrix will be 4 matrix of [3, 4], i.e. [3, 16].