Add `EncodedDepthImage` archetype

[oxkitsune]

DepthImages are currently handled as raw ImageBuffer inside, but there are use cases where depth images are stored in encoded formats, such as TIFF (e.g., LeRobot datasets).

Since TIFF decoding is implemented in #8970, this functionality can be reused to support encoded DepthImages.

[pablovela5620]

would something like this make it easier to compress depth maps so they don't take up as much memory (like rr.Image.compress() does)

[makeecat]

I have made a prototype working with re_mcap to visualize RVL encoded depth image from ros2 mcap bag.

The prototype has the following changes:

1. Implemented the EncodedDepthImage archetype
2. Introduced a new utility crate re_depth_compression to host rvl codecs implementation
3. Integrated EncodedDepthImage archetype into viewer
4. Supported the EncodedDepthImage archetype in re_mcap.

I would like to get advice from rerun team @oxkitsune on how to proceed merging this new feature into the codebase.

Here is my initial thought:

1. Submit a draft PR with a sample bag to show proof of concept so that rerun team can run the code and get a sense.
2. Get feedback on the general archetype design
3. Split the draft PR into four components: archetype, re_depth_completion crate, viewer integration, re_mcap integration

In the next step, we can implement or wrap more codecs in the re_depth_compression crate.

Please let me know your thoughts.

[oxkitsune]

I have made a prototype working with re_mcap to visualize RVL encoded depth image from ros2 mcap bag.

That is seriously awesome! Looking forward to it!

I would like to get advice from rerun team @oxkitsune on how to proceed merging this new feature into the codebase.

I think your suggestion of opening a draft PR is a good step, but I'd say that you can open the PR with the new archetype directly alongside it.

The other features (re_mcap etc) should be follow up PRs, agreed.

[makeecat]

The draft PR is proposed in #11877. The commits of PRs have separated each functionality