Multi-argument CV_Assert() is accessible via CV_Assert_N() (with malformed messages).

Enable Torch layers tests with Intel's Inference Engine backend.

* Custom deep learning layers support

* Stack custom deep learning layers

Main purpose of this namespace is to avoid using of incompatible
binaries that will cause applications crashes.

This additional namespace will not impact "Source code API".
This change allows to maintain ABI checks (with easy filtering out).

- winpack
- opencv_world

"Torch invalid argument 2: position must be smaller than LLONG_MAX"

These conditions are always true for "long position" argument.

https://github.com/opencv/opencv_contrib/tree/e6f63c7a38ca40c5dc33e38736e3027e3528d6cb/modules/dnn