add post_process_depth_estimation to image processors

[the-neural-networker]

What does this PR do?

This PR adds a post_process_depth_estimation method to the DPTImageProcessor class. The method takes the raw outputs of DPTForDepthEstimation and an optional target_sizes parameter, and returns a list of depth estimation maps resized to the specified target sizes (if provided) or in their original size. Since DepthAnything model also uses the DPTImageProcessor, this feature is applicable to DepthAnything as well.

Addresses #30917

Who can review?

@NielsRogge

[the-neural-networker]

Any suggestions are more than welcome!

[amyeroberts]

@the-neural-networker Thanks for working on this!

The main two requests would be:

• Add this for all depth estimation models
• Add tests to make sure the output is as expected