* Added set_dnn_prefer_smallest_algorithms()

* Added description for set_dnn_prefer_smallest_algorithms();w

* Changed architectures to algorithms.

runtime, even with very bad parameter selections.

- Moved all uses of PYBIND11_MAKE_OPAQUE to a common header to avoid possilbe ODR violations.
- Added python binding for make_bounding_box_regression_training_data().

of a tensor changed.  I.e. if you ran mini-batches with different sample sizes
you would get an assert triggering.  This has been fixed.

printed to cout or xml.