* Simplize Gradient Check

* Stash

* Extract apply_backward_pass to backward.py

Rename apply_backward_pass to append_backward_ops

* Use graph API to check gradient

* Fix ci

* Fix CI

* Fix backward for double precision

* Stash

* Fix CI

* Fix ci

* Ignore GRU test

* Ignore xe op

* Fix CI

* Fix softmax with xe gradient

The correct equation should be IG = OG * (d_softmax_with_xe())

* Fix typo

* Fix merge error

* Disable LRN

* compelete restoring program_bind from program_desc

* Fix bugs

* fix compile errors

* fix errors and add unit tests

* rename some vars

* Follow comments

* Feed/Fetch op just plain operator, not a OpWithKernel
* Do not register OpInfoMaker since Feed/Fetch will never be
  configured by users
* Feed/Fetch op has empty gradient
* Feed/Fetch op do not hard code `feed_variable`, `fetch_variable` as
  its input and output, make it as a plain Operator input/output

* Add debug string for Python ProtoBuf

and Rename `Sync` to `Flush`

* Add check of ProtoBuf initialization

1. InferShape should be a method for `OpDescBind` not `Operator`, since
  `OpDescBind` is a compile-time concept.
2. Pre-create operators and store them into a map. Make InferShape
   faster