* fix cond_op_test and add some check to operator.run

* tmp

* optimize kernel check

Add GradOpDescMaker to OpInfo and complete OperatorRegistrar method

use float32 in cond_op

Remove add_op since it can be replaced by sum_op

Design doc of compile time register gradient operators

Add Skeleton of Python API design

Add multiple Inputs/Outputs check interface

Add pool op

Update Python API Design

Using double precision to stablize lstm gradient check

* init remove cond_op infershape

* optimize code

* add PrepareDataForSubnet and MergeDataFromSubnet

Few typos in refactorization file

Unify Activation functions and simplify register code

Fix row buffer mem free

Fix infer when input is empty in v2/inference.py

refine "Testing Backward Operators" in new_op_cn/en.md

fix typo

remove rowwise_add_op