refactor(deviceio): make FrameMetadataTrackerOak single-stream

[jiwenc-nv]

Description

FrameMetadataTrackerOak held N SchemaTrackers, one per OAK stream, while every other tracker is 1:1 with a tensor collection — the one thing blocking it from #853's generated template. Narrowed to one stream per instance; create one tracker per stream. Supersedes #874.

The ctor takes a plain collection_id ("oak_camera/Color") rather than the (prefix, stream) in @shaosu-nvidia's comment: that pair is the generated template's exact signature, and every sibling tracker already takes a bare collection_id.

Breaking. get_stream_data/stream_count → get_data. Each tracker records under its own MCAP base name to oak/oak_tracked rather than one tracker writing a channel per stream, so McapRecordingConfig takes N entries and readers keyed on <base>/Color need updating. Two channels because the generated live template hardcodes reader indices 0, 1.

Out of scope. No replay impl — OAK never had one and codegen emits it. Manifest migration (last task of #868) is blocked on #853.

Refs #868

Type of change

• Breaking change (fix or feature that would cause existing functionality to change)

Testing

Built clean with cmake --preset py3.12 on Linux aarch64 (Orin); ctest -LE "gpu|window" passes 233/233 (the 4 failures in a full run are gpu/window-labelled viz tests, untouched here). Smoke-tested the wheel's new API surface.

No new automated tests: the existing OAK tests cover the schema, not the tracker, which needs a live OpenXR runtime plus hardware. examples/oxr/python/test_oak_camera.py --mode schema-pusher is updated but not yet run on an OAK-D.

Checklist

• I have read and understood the contribution guidelines
• I have run the linter and formatter with SKIP=check-copyright-year pre-commit run --all-files
• I have made corresponding changes to the documentation
• I have added tests that prove my fix/feature works (or explained why not)
• I have signed off all my commits (git commit -s) per the DCO
• Verified OAK-D end-to-end on hardware (task 5 of Refactor FrameMetadataTrackerOak to single-stream to fit codegen template #868)

[coderabbitai]

📝 Walkthrough

Walkthrough

OAK metadata tracking now creates one FrameMetadataTrackerOak per camera stream. The tracker accepts a collection ID and exposes get_data() without a stream index. The new oak_collection_id() helper validates stream types and builds collection identifiers for C++ and Python. Live tracking uses one schema reader and tracked-data object. Plugins, examples, recording configuration, and documentation now register and consume per-stream trackers.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• NVIDIA/IsaacTeleop#870: Updates device metadata-tracker architecture and recording integration for Orbbec trackers.
• NVIDIA/IsaacTeleop#874: Implements the same one-stream-per-FrameMetadataTrackerOak refactor across APIs, bindings, examples, and documentation.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 25.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: refactoring FrameMetadataTrackerOak to support a single stream.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch jiwenc-nv/IT868

---

_{Comment @coderabbitai help to get the list of available commands.}

[shaosu-nvidia]

Tested with test_oak_camera.py with device and confirmed the mcap is able to be written. Each tracker represents one stream and has both oak and oak_tracked channels written.

[jiwenc-nv]

@nv-jakob I hope this will unblock you to run the template refactor with the OAK-D stuff as well