Refine cmake for cudnn op

* Enhance bipartite_match_op to support argmax matching after bipartite matching.

* Fix typo error.

Refine the inference API and unittests

set the default option of WITH_FAST_BUNDLE_TEST be OFF

Fix send_recv unit test

Fix c-api doc style

* "try enhance concat op"

* "enhance the concat operator"

Fix the bug of conv2d 

fix short connection again

fix warp ctc dir error

add c-api quick start

Enhance layer_function_generator

create learning rate for multi program

move doc/api/v2/fluid to doc/api/fluid

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

move Fluid API doc/code out of V2 API doc/code