More truthful depth camera noise models

[peci1]

Desired behavior

It wouldn't hurt to implement some more noise models for depth cameras than the completely useless additive Gaussian noise.

I'm not saying I have time to implement them, but I wanted to open the issue to make a space for discussion and possible starting point for contributors who would want to work on that. Maybe next year's GSoC? :)

The first two models I propose are:

1. Multiplicative Gaussian (the noise is the larger the further the point is).
2. Sine-wave-like noise corresponding to the error induced by disparity matching in stereo: https://docs.luxonis.com/projects/hardware/en/latest/pages/guides/depth_accuracy .

Alternatives considered

One could also run a true stereo algorithm on two source cameras, but that would get quite computationally intensive. Moreover, users have the possibility to do that with the currently available implementation, they just need to choose and provide the stereo matching algorithm.

Implementation suggestion

Ideally, the noise models should be implemented as shaders.

The sine-wave-like noise should be parametrized by the camera resolution.

[azeey]

Related to gazebosim/gz-sim#2367