Enable h-label classification

[sungmanc]

Summary

This PR introduces the H-label Classification for OTX2.0. The final benchmark table will be uploaded soon.
Also, this PR will be edited after migrating #2758.

Notes,

• In this PR, the model metric is different from the OTX1.5. In the OTX1.5, the final metric is [top-1-acc. + mAP], however, I edited the metric to [top-1-acc + multilabel-acc] since it is more reasonable for me. Changing the model metric at the OTX side doesn't affect to the Geti.

The figure below represents the simple design of H-label data part

How to test

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have added e2e tests for validation.
• 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) 2023 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

[sungmanc]

@vinnamkim , thanks for the comments, I'll update this PR after merging this PR (#2758) since it brings the DataMetaInfo what I need.

[sovrasov]

When label resolving functionality is going to be added? As I see now, we have only metric computation, and therefore we don't need to produce human-readable predictions of the model, but in the future obtaining postprocessed results is also good to add (at least for Geti).

[sungmanc]

@jaegukhyun , @vinnamkim

This commit(fdc033e)

• introduces the HLabelMetaInfo and changed the location of DataMetaInfo. For the h-label classification, HLabelClsDataset need the information of h-label, however, in the previous implementation it is located at the DataModule side. That's the reason why I changed the location.
• Also I added the SubsetDataMetaInfo and DataMetaInfo to represent the information for all subsets (train, val, test). Currently, there are no cases that have different data classes between train and val. However, we need to care about the semi or self manner training.

[sungmanc]

When label resolving functionality is going to be added? As I see now, we have only metric computation, and therefore we don't need to produce human-readable predictions of the model, but in the future obtaining postprocessed results is also good to add (at least for Geti).

It seems great, I'll add that functionality at the later phase.

[vinnamkim]

LGTM.