1. Add option WITH_AMD_GPU.
2. Add cmake/hip.cmake for HIP toolchain.
3. Some external module such as eigen may need HIP port.
4. Add macro hip_library/hip_binary/hip_test to cmake/generic.cmake.
5. Add one HIP source concat.hip.cu as an example. Each .cu may have its corresponding .hip.cu.

* fix nccl op unit test

* fix build error

* format code

* refine nccl related unit test

* fix build error

* add setGPUData

* clean up

* follow comments

* rm test_nccl.cu

* follow comment

* rm wait

* compile and install the static library of fluid inference

* fix dynload_cuda not in CPU mode

* update shared library and adjust the deploy of openblas

* adjust the deploy of openblas

* * auto add all fluid modules for static library
* use libprotobuf.a instead of libprotobuf-lite.a for profiler

* use set_property to set the global varible instead of ENV

* add gpu depends of fluid modules, auto add inference_lib_dist depends

* change the condition of openblas_lib, and fix a typo

Enable whole-archive flag in cc_test and use cc_test to rewrite the CMakeLists.txt of inference unittest.

* fix grpc compile warn

* update

* -Wnon-virtual-dtor -> -Wno-non-virtual-dtor

* add gflags for C++ unittest

* WIP send recv op

* WIP send recv

* put grpc impl in details

* put grpc impl in details

* update wip

* update proto

* update proto

* update proto

* clean cmake

* wip on op implementations

* wip on op implementations

* compile ok adding ut

* wip unitest

* add extern cares for linking

* wip add ut

* working version send recv

* revert optimizer.py

* update test cmake

* add libtool to dockerfile

* update cmake dependency

* update cmake depends

* update cmake grpc depends

* fix cmake dependency

* fix compile error

* fix compile

* follow comments

* update

* update copyfrom

* py_test support argument, test_image_classification_train support argument

* use REMOVE_ITEM to rm item from list in cmake

* fix nv_library

* fix symbol in gpu_info.h

`merge_static_libs` also have the similar logic of using ${dummyfile},
I am not sure if there needs a change or not.

Python should be able to manipulate Protobuf message because:

1. Python's `create_op_creation_methods` take the `OpProto` array to
   generate all `op_creation_methods` in RunTime.

2. All `op_creation_methods` will create an `OpDesc` and pass it to
   Paddle C++ method `CreateOp` and return the Op handle.

Here is the list of what is added in this commit:

* Add `protobuf_generate_python` if it is not defined.
  * Before cmake 3.4, `protobuf_generate_python` is not defined. Just
    copy the implementation of that function in `protobuf.cmake`
* Add `py_proto_compile` function in `cmake/generic.cmake`.
  * It follows bazel's API interface.
    * https://github.com/pubref/rules_protobuf#rules
* Add an empty package named `paddle.v2.framework`, all python code of
  `paddle::framework` will be in that package.
* Generate protobuf's python module `__init__.py` by `touch` while
  compiling.
* Change setup.py.in, make `paddle.v2.framework.proto` uses the
  generated protobuf pythons.