fix(ci): skip the CUDA parity-gate default run on a build without CUDA

[lusoris]

Summary

Closes T-CI-PARITY-GATE-DEFAULT-RUN-NO-CUDA-BUILD-2026-10-01 (found while running --suite=fast --suite=gpu in a HIP-only build dir).

test_cuda_parity_gate_default_run (added in #1676) is registered in the slow and gpu suites of every build. On a libvmaf built without CUDA the CLI refuses --backend cuda (ADR-0498), every gate cell reports ERROR, and the test returned the gate's exit status, so a HIP-only or SYCL-only build dir had one failing test: 260 OK / 1 failed on ryzen-4090-arc with -Denable_hip=true -Denable_cuda=false.

The test already skipped when no CUDA device is present. A build without CUDA now skips the same way, through one helper (cuda_unavailable()). A CUDA build on a host whose device fails to initialise still fails.

core/test/test_cuda_parity_gate_skip.py (fast suite, device-free) builds the refusal from the format string in core/tools/vmaf.cpp and checks that it is a skip, that a parity failure is not, and that a missing SYCL backend says nothing about CUDA. It fails on the parent commit (3 errors).

Type

• fix — bug fix

Deep-dive deliverables (ADR-0108)

• Research digest — no digest needed: trivial (a skip condition in one test).
• Decision matrix — no alternatives: only-one-way fix.
• AGENTS.md invariant note — no rebase-sensitive invariants.
• Reproducer / smoke-test command — under "Reproducer" below.
• CHANGELOG fragment — changelog.d/fixed/ci-parity-gate-default-run-skip-no-cuda.md.
• Rebase note — no rebase impact: fork-only test files.

Reproducer

python3 core/test/test_cuda_parity_gate_skip.py
# with a binary built without CUDA:
VMAF_BUILD_DIR=build-hip python3 core/test/test_cuda_parity_gate_default_run.py; echo $?   # 77

Bug-status hygiene (ADR-0165)

• docs/state.md updated in this PR: T-CI-PARITY-GATE-DEFAULT-RUN-NO-CUDA-BUILD-2026-10-01 added to Recently closed.

Netflix golden-data gate (ADR-0024)

• I did not modify any assertAlmostEqual(...) score in the Netflix golden Python tests.