feat: Image segmentation for ios

[JakubGonera]

Description

Add semantic image segmentation with iOS native code. The result of a model run is a map containing masks of per-pixel probability for specified classes and an 'argmax' mask containing the indices of max value labels for each pixel.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update (improves or adds clarity to existing documentation)

Tested on

• iOS
• Android

Checklist

• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have updated the documentation accordingly
• My changes generate no new warnings

[chmjkb]

Overall, LGTM 🚀

[chmjkb]

Wondering if we should use this or keyof typeof DeeplabLabel. Using keyof typeof would make it consistent, since it is used in object detection but I think i like this one better. WDYT?

[JakubGonera]

I haven't notice the use in object detection, but I would still make the case for this. In object detection the enum is used as a return value, so we're just interested in the string content, whereas here the enum is supplied by the user (in both cases, either as the filter for classes in a list, or as the key to query results in the dict), so we care more about correctness. I think that autocompletion for the enum values when writing is worth more than consistency in this case.