coco_format: add an option to preserve original category IDs when extracting

[IRDonch]

Summary

This is needed for potential integration into Accuracy Checker, since AC has this option as well.

As a side effect, refactor the loading of categories to remove duplication between the panoptic segmentation and other task types.

How to test

Checklist

• I submit my changes into the develop branch
• I have added description of my changes into CHANGELOG
• I have updated the documentation accordingly
• I have added tests to cover my changes
• I have linked related issues

License

• I submit my code changes under the same MIT 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) 2021 Intel Corporation
#
# SPDX-License-Identifier: MIT

[zhiltsov-max]

I suggest to use more short "keep" instead of "preserve".

Update the format documentation.

[IRDonch]

I suggest to use more short "keep" instead of "preserve".

Okay, done.

Update the format documentation.

I would, but there's no documentation for the importer/extractor API in the first place.

[zhiltsov-max]

I would, but there's no documentation for the importer/extractor API in the first place.

How it hinders to describing import parameters in the format documentation? There is nothing about API at all.

[IRDonch]

The parameter currently only exists on the API level. Should I add a corresponding CLI option?

[zhiltsov-max]

IMO, there's a little sense, if it is an API-only parameter. Maybe, I'd invest in auto-generation of cli parsers from the ctor parameters, but not now.

[IRDonch]

OK, I added a CLI option equivalent and documented it.