Create Semantic Segmentation Model workflow block

[leeclemnet]

What does this PR do?

• Add Semantic Segmentation workflow block
• Post-process semseg response into sv.Detections for integration with mask visualizer using RLE encoding to serialize non-contiguous class masks (already being done for SAM3)
• Add support for visualizing RLE-encoded masks in MaskVisualizer (also benefits SAM3)

Type of Change

• New feature (non-breaking change that adds functionality)

Testing

• I have tested this change locally
• I have added/updated tests for this change
  • unit test
  • e2e integration test

Test details:

• unit test for semseg model workflow block
• integration test with mask viz
• test workflow UI in staging with local server

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code where necessary, particularly in hard-to-understand areas
• My changes generate no new warnings or errors
• I have updated the documentation accordingly (if applicable)

Additional Context

[PawelPeczek-Roboflow]

I believe output serialisation needs to be added for RLE encoded mask and following up on that - atm sv.Detections.from_inference(...) which we were alwasy recommending is not going to work with the result from server - leaving people with quite elaborated result they most likely don't know how to use
also - some e2e test would be nice

[leeclemnet]

TODO for future:

• PR to supervision to decode RLE-encoded masks (similar to how polygons are decoded) - enable sv.Detections.from_inference() to work on semseg outputs naturally
• Bump supervision version in inference
• For inference 2.0 consider RLE serialization for all masks, no polygon conversion