* add interface and unittest of RecurrentGradientMachine for the function of multiple Subsequence inlinks with unequal token length

* Update readme

* Apache 2.0

* Update Readme

* Update readme

* Update readme

* Rerange Build docs & emphasize them in README.md

* Rerange Build docs & emphasize them in README.md

* Add automatic check AVX in CMake

* Add indent in FindAVX.cmake

* Revise table format and some words in build docs

* Update build docs

* Update build docs

* some bug fix for sparse matrix

* a minor bug fix

* also refine install docs, too

In some situation, for instance, in the virtual machine, it could happen.

* Add automatic check AVX in CMake

* Add indent in FindAVX.cmake

* Revise table format and some words in build docs

* Update build docs

Fix cmake/FindAVX.cmake

* Add automatic check AVX in CMake
* Revise table format and some words in build docs

Add Mac OS X port

* follow comments to fix bugs

Remove main function in some unittest.

* refine sparse momentum api and unittest
* fix unittests bug

This bug occasionally causes dead lock in test_RecurrentGradientMachine
In general, conditional_variable::notify should be used together with mutex for changing condition.

* Add noavx to docker
* Not to use directory to split Docker image.

Add analystic scripts to documentation

* reduce trainer count for unit test on MAC OSX

* Fix lots of trainer_config_helpers bug, and complete unittest for `layers.py`

* Correctly handling multiple inputs and integer inputs for recurrent_group

* Fix ScatterAgentLayer for generation

* Revert sequence_(nest)_rnn.conf

* Add `device` parameter to ExtraAttr in trainer_config_helpers.
* add unittest for it.

* add type compatible check for ParamAttr

Local training with "sparse_update = True" parameter triggers kSgdSparseCpuTraining mode, fix bugs under it.