test(inference,#1275): regression test for no-CPU-fallback alpha contract

[joelteply]

Summary

Add tests/no_cpu_fallback_contract.rs — three forbidden-strings ratchets that fail the build if a future PR weakens the no-CPU-fallback alpha contract:

1. select_best_device_panics_loudly_on_no_gpu — asserts inference/model.rs::select_best_device keeps the `panic!("No GPU device available for inference. CPU fallback is disabled.")` loud-fail and tries CUDA + Metal before panicking.
2. ort_providers_documents_no_cpu_fallback_contract — asserts ort_providers.rs keeps the "CPU fallback is forbidden" comment that documents the rule from source.
3. llamacpp_adapter_uses_loud_fail_for_no_local_model — asserts LlamaCppAdapter uses the typed NoLocalModelLoadable error (shipped in docs(vdd): record RTX Qwen2.5-Omni result #1093 / lane A PR-2) rather than a silent skip.

Why

Without a regression test, the no-CPU-fallback contract is invariant by inspection only. A test makes it invariant by construction; future PRs that want to reintroduce a fallback (the audit at #1262 found three separate dead-code remnants of past attempts) hit the test rather than ship.

Pattern: same forbidden-strings ratchet shape as lane F PR-2 (#1129 — TS persona forbidden-strings), applied to the Rust inference layer. A test failure points the future-PR-author at the exact contract they're about to weaken.

Audit context

Final PR (4 of 4) for #1262 (silent CPU model fallbacks audit). Closes acceptance criterion #3 of #1262 ("regression test per fallback path").

Sibling PRs:

• fix(inference,#1262): delete dead compute_router.rs (no-CPU-fallback alpha) #1277 — delete dead compute_router.rs (213 LOC)
• fix(inference,#1273): delete dead Candle Qwen3.5 GGUF backend (~1100 LOC) #1279 — delete dead Candle Qwen3.5 path (1132 LOC)
• fix(inference,#1274): delete dead vendored/metal_deltanet.rs (+ shader) #1281 — delete dead vendored/metal_deltanet.rs (164 LOC)

Audit: #1262 (comment)
Lane: alpha flywheel #1272 lane 6.

Test plan

• cargo test --features metal --test no_cpu_fallback_contract — 3 passed, 0 failed
• Precommit (TypeScript + browser ping): PASSED

🤖 Generated with Claude Code