add clear_gradient、current_step_lr api and parameter_list parameter for optimizer in dygraph (#1694)

* add clear_gradient function and parameter_list parameter for optimizer, test=develop

* fix AdagradOptimizer_cn.rst, test=develop

* add sample code, test=develop

* polish sample code, test=develop

* polish description for clear_gradient, test=develop

* add current_step_lr api for optimizer, test=develop

* polish backward api doc test=develop, test=document_preview

* polish backward api doc test=develop, test=document_preview

* polish doc of optimizer test=develop, test=document_preview

* polish doc of optimizer test=develop, test=document_preview

* polish doc of optimizer test=develop, test=document_preview

* polish code of no_grad_set test=develop

* Fix problems of chinese docs

* Revise chinese docs

* Add more fixes

* Add data type comments for some inputs