ci: Copy nightly matrix into override to test GH UI loading

[vyasr]

Problem

The nightly CI workflow has become too large for the GitHub UI to load/render.

Approach

Copy the full nightly matrix (65 entries) into the workflows.override section so it runs as a PR workflow. This lets us observe whether the GH UI can handle the nightly job count, and iteratively reduce it to find a manageable subset.

Important Notes

• This PR is intentionally non-mergeable — the override section blocks merging by design.
• The actual nightly workflow is unaffected (it doesn't use --allow-override).
• Once a working subset is identified (or the diagnosis is complete), the override will be reset to empty.
• Commit uses [skip-vdc][skip-docs][skip-tpt] to reduce unrelated CI noise during this experiment.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: dce940fd-4d98-4c1d-98a6-186033ca0df7

📥 Commits

Reviewing files that changed from the base of the PR and between b01ac23 and 708c815.

📒 Files selected for processing (1)

• ci/matrix.yaml

🚧 Files skipped from review as they are similar to previous changes (1)

• ci/matrix.yaml

---

📝 Walkthrough

Summary by CodeRabbit

• Chores
  • Expanded continuous integration to broaden build and test coverage across multiple CTK versions, GPU targets and architectures (including arm64), add GPU-specific checks, packaging and install tests, Python test entries, compiler/tooling builds and linting, and additional architecture-specific build groups to improve overall platform reliability.

suggestion:

Walkthrough

The PR expands ci/matrix.yaml's workflows.override matrix from a single entry to comprehensive enumeration of CI job configurations. Coverage now spans CTK versions 12.0, 12.X, 13.0, and 13.X across default projects and cudax builds, multi-target GPU testing with limited RTX PRO 6000 access, specialized configurations for arm64/thrust/NVRTC/C bindings, Python testing, packaging validation, and compiler-specific builds (NVHPC, clang-cuda, clang-tidy) with architecture-specific sm grouping.

Changes

CI Matrix Expansion for CTK and Compiler Coverage

Layer / File(s)	Summary
Build matrix for default projects and cudax ci/matrix.yaml	Expanded build matrix entries for CTK 12.0, 12.X, 13.0, and 13.X targeting both default projects and cudax, each with compiler/standard groupings.
GPU testing matrices with hardware coverage ci/matrix.yaml	CTK 12.X and 13.X testing matrices covering multi-project GPU tests and limited RTX PRO 6000 target coverage via test_nolid and test_lid0 variants.
Specialized test and build configurations ci/matrix.yaml	Arm64 build coverage, thrust GPU dispatch mode checks, NVRTC/verify_codegen jobs with CTK support constraints, and cccl_c_parallel/cccl_c_stf pinned matrices with limited RTX PRO 6000 coverage.
Python test and packaging matrix additions ci/matrix.yaml	Python test override entries with Linux-consistency CTK/py_version/GPU/compiler configurations, plus CCCL packaging test/install entries and NVBench helper test matrices across CTK and host compiler combinations.
Compiler-specific and architecture-focused builds ci/matrix.yaml	NVHPC, clang-cuda (with sm list coverage), clang-tidy, and architecture/family-specific sm build group entries for selected projects.

---

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

[jrhemstad]

/ok to test

[jrhemstad]

The StepSecurity reports should be disabled now. Kicked off another run to see if that fixes the issue.

[jrhemstad]

/ok to test 708c815

[github-actions]

😬 CI Workflow Results

🟥 Finished in 1h 36m: Pass: 90%/1054 | Total: 8d 20h | Max: 58m 58s | Hits: 99%/1196940

See results here.