Fix learning rate scheduler performance issue

Feature/polish reshape and lod tensor blocking queue

* Add new_doc to develop branch

* Add pics of Anakin tutorial

* Update anakin_tutorial.md

* Delete anakin_fm_en.png

* Delete contri1.JPG

* Delete contri2.JPG

* Update build_and_install_lib_cn.rst

* Update compile_paddle_lib_cn.md

* Update compile_paddle_lib_en.md

* Update install_doc.rst

* Update install_doc.rst

add profiler for demo_trainer

Change latest to 0.0.0

Modify PyReader doc in python/paddle/fluid/layers/io.py

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into feature/polish_reshape_and_lod_tensor_blocking_queue

* expose rank_loss python api

Speedup reduce sum grad op

Skip reader when save

Speedup reduce sum grad op

Some enhancement on readers

disable warning

* get diff for all changes in current branch

* diff with upstream/develop

* let dist test serial

1. Make the feeding thread of py_reader a daemon thread.
2. Update buffer_reader's destructor, fixing a bug.
3. Make pyreader demo script supporting CPU environment.

* Fix serious bug in nesterov momentum optimizer.

Fix .sh file

* wip add se-resnext test for 2x2 and parallel exe

* testing

* finish dist unitest 2x2

* update python interp

* fix gpu number

* make this test can run in parallel

Change and polish readers

* Fix Reduce and Gather

* Fix unit test