* Support testing during training by ParallelExecutor.

* Add unit test.

* Improve the interface.

* Follow comments.

* Update source files.

* Update headers

* Update

* Update

* Update

* Update

* Fix a CMake dependency

* Fix cpplint issues

* Fix cpplint issues in channel.h and channel_impl.h

* Fix typo

All leaves must be variables. When all variables are ready, the
execution will be completed. If a operator has no output, the `Op::Run`
might not be started when the execution of graph has been complete.

* fix data transform when op have inplace in/out

* add log

* should not delete scope because Compute maybe async

* optimize code

* Fix Issue 9388

* Fix typos