Make users can access parameters of layers and their gradients.

* Add squared error layers doc

* Add doc for gru_unit

* Remove cdot which isn't supported

* Update layers.rst

* Update layers.rst (minor)

* "remove cudnndevicecontext"

* "remove unused init code"

* "fix hash functions"

* update docs

* follow comments

* fix typo

* "change operator interface"

* "move devicepool to device_context"

* "fix operator test"

* "fix op_registry Run interface"

* "net op passed. Need to fix nccl multi-Context"

* "add nccl group function"

* "add nccl group function"

* "fix gpu count exceed 32 error"

* "fix recurrent op, nccl op"

* "change the other operators interface with Place"

* "fix typo"

* "fix pybind"

* "fix device in python side"

* "fix pybind failed"

* "add init for test"

* "fix CI"

* add design doc on keys of operator kernel type

* follow comments