CPU optimization integration checks (2026-09-17)

These are correctness/build logs, not new timing measurements.

• strided-rs PR259 merged as 5bc5ab75a20277f0c8820cb288b23f6bb6dfbd91.
• tenferro-rs PR1807 source 6ac70cd07e802dee3bda969bc40e0ffd91f37168; its follow-up c184115e updates only the CI test's expected dependency hash.
• Docker image sha256:228bf2e4baf4ef4117a65bb8a5a2d58247ce0c0985aec8a8fb53b7a58193893f, Rust1.98.1, shared OpenBLAS0.3.26. All six strided crates resolve from GitHub, without Cargo path/manifest overrides. Runtime OpenBLAS/OMP/Rayon threads1; Cargo jobs16 and Rust test harness threads1.

tenferro-pr-fast.log.gz: passed formatting, doc snippets, strict workspace Clippy plus tropical/sparse/TBLIS extension Clippy, and 1439 tests/doctests:

bash scripts/check-pr-fast.sh --base origin/main --coverage-reviewed \
  --test 'cargo test -j 16 -p tenferro-cpu -p tenferro-ad --no-fail-fast -- --test-threads=1'

tenferro-pr-blas.log.gz: six BLAS-only tests and one injected-provider opt-out:

export RUSTFLAGS='-L native=/opt/openblas/lib -l openblas'
cargo test -j 16 -p tenferro-cpu --no-default-features --features cpu-blas \
  --lib blas_uninit_tests -- --test-threads=1
cargo test -j 16 -p tenferro-cpu --no-default-features \
  --features cpu-blas,provider-inject --lib injected_blas_has_no_uninit_witness \
  -- --test-threads=1

tenferro-ci-config.log.gz: python3 scripts/ci/run_profile.py ci-config passed all328 Python checks and actionlint after updating the expected pin. Provisioning messages in these unit tests are mocked; no cloud resources were provisioned by this local check.

These gates do not establish native performance acceptance or the deferred4T matrix. The benchmark's full timed linalg-AD integration script was not rerun.