1. Add attr 'average' into ParamAttr.
2. Make 'params_grads' optional for AverageModel.
3. Add option 'average_mean' and 'average_variance' for batch_normal.

In this way we can build and test using WITH_FLUID_ONLY flag being set
to ON.

- move paddle.v2.dataset,reader to paddle.dataset,reader
- remove unused code (which depends on v2) in paddle.dataset,reader

Fix dist error with lr decay layer

This reverts commit ce16400d.

* "add functors"

* "remove old code"

* "fix"

* "fix ci"

* "add details"

* "fix ci"

* "fix ci"

* "fix ci"

* "fix ci"

* "remove unused code"

This is a temporary solution to allowing for variables to be copied during a channel send operations.  Also fixed issue with is_copy for "channel_send" method, and also updated unit tests.

* Fluid Channel should match Go Channel in Semantics

* Fix Python channel_send

* Address code rveiew feedback

* Fix open_files_op.cc

* Add description to Channel Asserts

* add split_ids_op

* add TestSplitIdsOp

* fix comment

* add test for empty tensor

* clean code

* rm unused code

* add LRN layer for fluid

* add LRN layer for fluid

* add documentation for LRN layer

* add paper reference for LRN layer

* add seperate documentation for LRN layer

* rm lrn.py in doc/fluid/dev/src

* change code style in lrn

* fix style of comments in lrn

* support empty tensor

* Move Select to concurrency.py; incorporate outputs

* CLang formatting for concurrency

* Remove extra bracket - formatting fix - 3

* Comment fix

* Temporarily fix bug for backward tanspiler when using parallel_do operator.

* Fix bug for backward tanspiler when using parallel_do operator