feat: Add getitrack video I/O, annotators, and detector adapters

[omkar-334]

Adds the video I/O, visualization, and detector-adapter layer for getitrack.

Included

• io/: VideoReader and VideoWriter, OpenCV context managers over BGR uint8 frames.
• visualization/: TrackAnnotator (draws boxes, ids, class names, and scores) and VideoAnnotator (annotate and write straight to a video file). Track colors are deterministic per id.
• adapters/: a DetectionAdapter interface and GetiAdapter, which wraps a getitune detector into getitrack Detections. getitune and torch are imported lazily, so getitrack installs without them.

Dependencies

• Builds on the core PR for Detections / TrackedDetections. This branch is stacked on getitrack-core-v2, so the diff includes the core changes until that PR merges into feature/getitrack.
• The CLI and public-API exports are not part of this PR; they depend on the ByteTrack PR and follow separately.

Testing

• Unit tests for the readers/writers, annotators, and adapters. Adapter tests that need getitune are skipped when it is not installed.

[omkar-334]

need to rebase this once #6796 is merged. This PR contains commits from that pr

[kprokofi]

@omkar-334, please, rebase

[kprokofi]

Overall, the structure looks good. However, as Leo already mentioned, typing is done badly. Could you work to type your variables / entities properly? Try to avoid using noqa/ignore if possible

[Copilot]

Pull request overview

This PR adds the “integration layer” around getitrack’s core detection/track types: OpenCV-backed video I/O, utilities for drawing tracked detections onto frames and emitting annotated videos, and a detector-adapter interface plus a getitune-backed adapter (with lazy imports so getitrack can be installed without torch/getitune).

Changes:

• Added VideoReader / VideoWriter context managers for iterating and writing BGR uint8 frames via OpenCV.
• Added TrackAnnotator / VideoAnnotator + deterministic per-track coloring for visualization.
• Added DetectionAdapter interface and a GetiAdapter to convert getitune outputs into getitrack Detections, plus unit tests (skipping getitune/torch-dependent tests when not installed).

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
libraries/getitrack/src/getitrack/io/video.py	Implements OpenCV-backed VideoReader/VideoWriter context managers and frame validation.
libraries/getitrack/src/getitrack/io/init.py	Exposes video I/O types from the getitrack.io package.
libraries/getitrack/src/getitrack/visualization/annotator.py	Implements drawing of tracked detections and a writer-composition helper (VideoAnnotator).
libraries/getitrack/src/getitrack/visualization/init.py	Exposes visualization utilities from the getitrack.visualization package.
libraries/getitrack/src/getitrack/adapters/base.py	Defines the DetectionAdapter interface for framework-agnostic detection.
libraries/getitrack/src/getitrack/adapters/geti.py	Adds GetiAdapter with lazy torch/getitune usage and batch→Detections conversion.
libraries/getitrack/src/getitrack/adapters/init.py	Exposes adapter types from the getitrack.adapters package.
libraries/getitrack/tests/unit/test_video_io.py	Unit tests for video I/O roundtrips, metadata, and error cases.
libraries/getitrack/tests/unit/test_annotator.py	Unit tests for annotator behavior, labels, colors, and video annotation output.
libraries/getitrack/tests/unit/test_adapters.py	Unit tests for adapter conversions and optional getitune/torch-dependent preprocessing/detect.