Fix for program crash when destructor is called before channel close with blocked readers/writers (#8197)

* Fix destructor crash and add unit tests

* Fix typo in unit test

* Reword comments

* Make close channel a generic test

* Refactoring unit tests

* Fix method name

* Fix parallel.do with batch norm

* Change log level

* CopyShare AllPlaces

* disable nccl test

* bring back parallel_do test

Impl scalar switch case op with condition op

* set up python code

* fix bug

* add cc file

* fix cmake

* add inference test for label semantic role

* fix

* address comments

* address comments

* address comments

* address comments

* add use_cuda

* Adding panic logic and test case

* Change panic behavior to boolean instead of exception

* Adding atomic

* Switch to boolean

* Fix spacing

* Add to close method

* Add test case to return zero on a closed channel

* Rename method

* Fix test

* ReceiveFromBufferedChannelReturnResidualValuesTest

* Adding the variable and case for unbuffered channel

* Fix review comments

* Fix format

* Remove a zero-value comparison

* First basic implementation

* Add infer example for image_classification

* Address review comments: round 1

Change the behaviour of thread pool. Thread pool will ignore the
thrown exception implicitly. It is hard to debug.
Corrently, ThreadPool::Run will invoke `LOG(FATAL)` if an exception
thrown.

Add a new type `Reader` for `VarDesc`, which can holds more than one
LoDTensor.