fix: port HoloHub Orin GPU detection

[wyli]

Summary

• Port the HoloHub CLI fix from CLI: Fix GPU_TYPE detection for Jetson AGX Orin on JetPack 7.x holohub#1601 for Orin GPU_TYPE detection.
• Classify Orin (nvgpu) as igpu for CUDA 12 / JetPack 6.x and dgpu for CUDA 13 / JetPack 7.x.
• Add SDK utility tests for Orin CUDA-major disambiguation and non-Orin fallback behavior.

Upstream CLI Audit

• HoloHub CLI commits after the last local sync through #1596: #1597 and #1601.
• #1597 package --cuda forwarding is already present in this repo.
• #1601 is ported in this PR.
• Other recent upstream commits were app/module/docs/website changes, not holoscan-cli changes.

Validation

• python -m pytest -q tests/unit/test_sdk_utils.py tests/unit/test_package_cmd.py
• python -m pytest -q
• python -m ruff check src/holoscan_cli/utils/sdk.py tests/unit/test_sdk_utils.py

AI-assisted: Created with Codex/GPT at the user's request.

Summary by CodeRabbit

Release Notes

• Bug Fixes

  • Enhanced GPU type detection on Orin-based systems to accurately distinguish between integrated and discrete GPUs based on detected CUDA version.

• Documentation

  • Updated GPU identification and CUDA container tag mapping documentation for improved clarity.

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: da48a54f-b778-4d09-8153-45316eac4441

📥 Commits

Reviewing files that changed from the base of the PR and between bf84ca4 and ae6019a.

📒 Files selected for processing (2)

• src/holoscan_cli/utils/sdk.py
• tests/unit/test_sdk_utils.py

---

Walkthrough

This PR updates GPU type detection for Orin systems and supporting documentation. The get_host_gpu() function now distinguishes between integrated and discrete GPUs based on the detected CUDA version, replacing the prior unconditional integrated GPU return for Orin hosts. Inline documentation for CUDA container tags is revised, and test coverage is expanded to validate the new branching logic.

Changes

Orin GPU Type Detection Refinement

Layer / File(s)	Summary
GPU detection logic and documentation src/holoscan_cli/utils/sdk.py	get_host_gpu() now returns igpu for Orin systems only when CUDA version 12 is detected; otherwise returns dgpu. CUDA container tag docstring updated to reflect revised JP/IGX OS associations for cuda13 and cuda12-igpu mappings.
Test coverage for GPU detection tests/unit/test_sdk_utils.py	Test suite replaces single Orin case with scenario-based validation: Orin + CUDA 12 → igpu, Orin + CUDA 13 → dgpu, Orin + no CUDA → dgpu, non-Orin → dgpu. Cache is cleared between scenarios.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Suggested reviewers

• MMelQin
• tbirdso

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix: port HoloHub Orin GPU detection' directly and clearly summarizes the main change: porting a fix for Orin GPU type detection logic from HoloHub to holoscan-cli.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

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