Speed up int scanner of data converter.

* Pre-allocated buffer for int scanner

Using linb::any/std::any instead of FunctionConfig

Refine NCE Layer

* Also test std::vector
* Use std::vector to PadConf

Add extra_layers in paddle.trainer.SGD.

fix parameter shape in v2

fix a bug of AgentLayer.

RecurrentLayerGroup for training.

Add the check of image size in exconv.

Feature/fix evaluator in layer v2

add wmt14 pretrained model

Fix hsigmoid layer comments.

Default `--rm` when `docker run` in README.md

Otherwise, it will create many containers which are not usefull.

Disable sparse_remote_update when local in paddle.v2

doc_cn mathjax_path 使用国内镜像以加快访问速度

Add plot to whl package

refine plot_curve