[Datumaro] Diff with exact annotation matching

[zhiltsov-max]

Motivation and context

It can be hard to find tools for direct equality comparison of datasets. This patch adds this capability to Datumaro.

• Added project ediff command for direct comparison of datasets
  • Optional annotation fields can be excluded from comparison (e.g. label for boxes, id, group)
  • Specific annotation and item attributes can be excluded from comparison (e.g. frame, is_crowd etc.)

Example output:

# Compare 2 projects, 
# ignore annotation .id and .group fields and the 'is_crowd' annotation attribute
# ignore item's 'frame' and 'id' attributes
# Result: compared CVAT xml and COCO instance annotations
datum project ediff test_project2 -if id -if group -ia is_crowd -iia frame -iia id

Found:
The first project has 0 unmatched items
The second project has 0 unmatched items
25 item conflicts
15 matching annotations
0 mismatching annotations

Output has been saved to 'diff.json' # has a list of differences

How to test:

1. Create a pair of datumaro projects (import a CVAT task etc.)
2. datum project ediff -p project1/ project2/

How has this been tested?

Unit tests

Checklist

• I submit my changes into the develop branch
• I have added description of my changes into CHANGELOG file
• I have updated the documentation accordingly
• I have added tests to cover my changes
• I have linked related issues (read github docs)
• I have increased versions of npm packages if it is necessary (cvat-canvas,
  cvat-core, cvat-data and cvat-ui)

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) 2020 Intel Corporation
#
# SPDX-License-Identifier: MIT

[zhiltsov-max]

Ready for testing.

[coveralls]

Pull Request Test Coverage Report for Build 7260

• 196 of 278 (70.5%) changed or added relevant lines in 5 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.05%) to 69.98%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
datumaro/datumaro/cli/contexts/project/diff.py	0	1	0.0%
datumaro/datumaro/components/extractor.py	3	4	75.0%
datumaro/datumaro/cli/contexts/project/init.py	9	40	22.5%
datumaro/datumaro/components/operations.py	182	231	78.79%

Totals
Change from base Build 7253:	0.05%
Covered Lines:	12119
Relevant Lines:	16868

---

💛 - Coveralls

[nmanovic]

ediff -> compare?

[nmanovic]

I would rename ediff to compare or something like that.