Standardize confidence_map

[Victorlouisdg]

Describe the feature you'd like
To get point cloud without long trail around objects with the ZED2i cameras, I rely on the confidence maps they provide for filtering. Currently these are accessible in airo-mono via the non-interface Zed2i function _retrieve_confidence_map(). For this reason, they are currently also not included in the MultiprocessRGBDReceiver. However I need the confidence maps and need to run my camera in a separate process for video recording, so it would be great if confidence maps were added.

Use cases
Anyone who want to filter low-confidence points from their point cloud.

Possible implementation
A DepthConfidenceMapType with a standardized semantic meaning. The ZED SDK returns float values between 0.0 and 100.0 where higher values mean more uncertainty, e.g. around the edges of objects. The Realsense SDK also has something based on disparity I believe: IntelRealSense/librealsense#3185

Additionally we make get_confidence_map() an (optional) interface function, and add it to the Multiprocess classes.