Document TensorRT Workflow.

Improvements to TensorRT Pipeline.

* Reverse throttle and steering.
* Reduce the number of nodes in the frozen graph further.

Test: Tested with donkey tubplot.

crop and normalize outside of model

Support for TensorRTLinear model.

changed Raspbian link from "Latest" to "Raspbian Stretch Lite"

* This PR adds support for Tensor RT.
* There are probably more minor improvements I can make to
`tensorrt.py`, but I think this is a good starting point.
* `freeze_model.py` exports metadata about the model which helps us
reconstruct the TensorRT model.
* Make sure `pycuda.autoinit` is never loaded during training, as it
interferes with TF-GPU (when training).

Test: Tested this on my Jetson Nano.

Improvements to the ML Pipeline

Add support for freezing Keras Models. * Add freeze_model.py to convert to a frozen TF graph. * Use `convert-to-uff` to convert the frozen graph to UFF. * Profit.

CSI camera: Config option to flip the cam frame orientation

If the camera is mounted in a rotated position, provide an option
to flip the camera frame.

CSICamera Jetson nano part is implemented using gstreamer. Passing
the flip parameter to the gstreamer pipeline will do this.

Signed-off-by: Jithu Joseph <jithu83@gmail.com>

Minor update to RotaryEncoder in encoder.py part (fix shutdown code)

Add axis names and button names to Joystick.

Fixes issue where images come out as BGR instead of RGB for Jeston Nano