Conversion to depth image does not work with animated focal_length

The current implementation assumes that the focal length of the camera is constant for all frames, which will lead to wrong results if it is animated.

No file, test, or entry point is named. Start by locating the depth-image conversion and its handling of focal_length, then inspect how per-frame camera values are represented. Done means animated focal_length produces correct depth images for every frame, with regression coverage if the repository has relevant tests.

Written by the indexing model from the issue text.