Semantic DVS Camera

[qhaas]

This is a feature request that has a workaround.

It would be useful for machine learning purposes to have the semantic tag associated with a given DVSEvent from the DVS sensor, analogous to how the LIDAR sensor has a semantic lidar variant.

We currently plan to assume that the tag should match the semantic segmentation camera output in the same tick, but that incurs overhead if we are not already capturing with the semantic segmentation camera or if we have (relative to the observer) fast moving objects in the scene where it isn't clear after the DVS camera events have been generated via interpolation between diffs of slower frame rate captures if the label should be the current or prior frame.