Update visual prompting pipeline for multi-label zero-shot learning support

[sungchul2]

Summary

This PR includes below updates for multi-label support:

• Add ZeroShotPromptType
• Migrate polygons or bitmap masks into prompts
• Integrate inference_step

** The table below shows reproduced results based on dice score.

• There is a performance difference between before and after this update because:
  • polygon and bitmap mask that were ground truth are used as prompts as well (bbox/point -> bbox/point/polygon/bitmap mask)
  • predicted masks are not multiplied with scores anymore
  • after multi-label enablement, benchmark performance can be changed one more because the current version has only considered single-label inference

model	task	before (52fef3d)	after (a811a11)
sam_tiny_vit (fine-tuning)	test	0.91	0.92
	export	1.00	0.92
	optimize	0.99	0.92
sam_vit_b (fine-tuning)	test	0.93	0.94
	export	1.00	0.94
	optimize	1.00	0.93
sam_tiny_vit (zero-shot)	test	0.59	0.53
	export	0.59	0.58
	optimize	0.55	0.52
sam_vit_b (zero-shot)	test	0.32	0.29
	export	0.33	0.30
	optimize	0.29	0.26

How to test

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have ran e2e tests and there is no issues.
• I have added the description of my changes into CHANGELOG in my target branch (e.g., CHANGELOG in develop).​
• I have updated the documentation in my target branch accordingly (e.g., documentation in develop).
• I have linked related issues.

License

• I submit my code changes under the same Apache License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below).

# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

[sovrasov]

@sungchul2 could you provide ZSL results on reduced coco (torch+OV) to prove there's no regression?

[eunwoosh]

Thanks for your work :) I left some comments. Please take a look.

[kprokofi]

postpone to merge PR before 2.2/2.3 is finalized