By shell command

```bash
sed -i 's#ifdef PADDLE_ONLY_CPU#ifndef PADDLE_WITH_GPU#g' `find ./paddle/ -name '*.h' -o -name '*.cc' -o -name '*.cpp' -o -name '*.c' -o -name '*.cu'`
sed -i 's#ifndef PADDLE_ONLY_CPU#ifdef PADDLE_WITH_GPU#g' `find ./paddle/ -name '*.h' -o -name '*.cc' -o -name '*.cpp' -o -name '*.c' -o -name '*.cu'`
```

Also remove operator bool in Error. The Error should be removed
later because it is not necessary for Paddle. We are now using Enforce
to handle error.

Use CHECK instead of PCHECK, because PCHECK is used for errno.

The implementation of BarrierStat is buggy, and it is not necessary
for Paddle to diagnose which node in cluster is slow.

Fix #2567

__must_check is a macro mark of function return value. It let developer
must check the return value is legal or not.

1. v2.layer.parse_network does not correctly handle the generation output.
2. GatherAgentLayer does not correctly handle generation output when batch_size > 1
3. Fix CustomStackTrace for rnn group

This only fixes the issue for ATLAS
MKL is still another fix.

In the original util.cmake. enable_virtualenv.c is always regenerated when cmake files are changed, which leads to the relinking of all the targets depends on paddle_utils.

* set PYTHON_HOME outside the paddle processes.

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

 using the same implemetation as mac os.