fix(ci): make DeviceOffloadPolicy Default derivable (clippy -D warnings)

[justinchuby]

CI hotfix — make DeviceOffloadPolicy Default derivable (clippy -D warnings)

Unblocks origin/main — the "CUDA compile (Linux+Windows)" job is RED on every PR (including #574) due to a clippy::derivable_impls failure under -D warnings.

Root cause

#572 flipped DeviceOffloadPolicy::async_pagein's default to false. That made the hand-written impl Default fully equal to a derivable default (all three fields = their type default: bool → false, Option → None), so clippy::derivable_impls now fires at crates/onnx-runtime-ep-cuda/src/weight_paging.rs:117.

Fix (provably behavior-identical)

Replace the hand-written impl Default for DeviceOffloadPolicy with #[derive(Default)] on the struct. The derive produces exactly the same defaults — enabled: false, device_budget_bytes: None, async_pagein: false. from_env() and all runtime behavior unchanged; field doc comments preserved. Net diff: 1 insertion, 11 deletions, one file.

Green confirmation

• cargo clippy -p onnx-runtime-ep-cuda --features cuda -- -D warnings → exit 0 (the failing gate; no other derivable_impls / -D warnings hits in the crate).
• CI first step cargo check --locked -p onnx-runtime-ep-cuda -p onnx-runtime-python --features onnx-runtime-python/cuda → exit 0.
• cargo fmt --all --check → clean.
• Existing guard test weight_paging::tests::device_policy_defaults_to_disabled (asserts enabled=false, device_budget_bytes=None, async_pagein=false) → passes, locking the derived defaults against field default-drift. No new test needed.

Decision note: .squad/decisions/inbox/cohaagen-cuda-clippy-hotfix.md.

Trivial lint hotfix — fast-track to unblock main.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.51%. Comparing base (03f9e60) to head (175e972).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #575      +/-   ##
==========================================
+ Coverage   80.75%   81.51%   +0.76%     
==========================================
  Files         315      315              
  Lines      123737   123737              
  Branches   123737   123737              
==========================================
+ Hits        99920   100869     +949     
+ Misses      19746    18791     -955     
- Partials     4071     4077       +6     

Flag	Coverage Δ
cli-ort-linux	86.69% <ø> (ø)
cli-ort-windows	83.84% <ø> (+0.10%)	⬆️
mlas	78.72% <ø> (+0.81%)	⬆️
offline	81.34% <ø> (+0.79%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 7 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

⚠️ Benchmark Change Detected

Comparison of criterion micro-benchmarks: PR head vs merge-base, measured on the same runner in the same job (base first → PR second).

ℹ️ Absolute times are informational only — they vary with runner load. The % change column is the reliable signal because both sides ran under identical conditions.

Status	Scenario	Base	PR	Change
⚠️	gather/large_f16_threads=1-internal/131072	13.87 µs	17.48 µs	+26.0%
⚠️	gather/large_bf16_threads=1-internal/131072	12.54 µs	14.86 µs	+18.5%
✅	logit_processing/seven_processor_chain_per_step	1.35 ms	1.51 ms	+12.5%
✅	matmul/medium_generic_bf16_threads=8/32x512x512	651.39 µs	729.75 µs	+12.0%
✅	matmul/large_generic_f32_threads=8/32x1024x1024	6.30 ms	7.01 ms	+11.1%
✅	tokenization/encode_tokens_per_second	424.70 µs	468.86 µs	+10.4%
✅	matmul/small_generic_bf16_threads=8/1x256x256	41.69 µs	45.37 µs	+8.8%
✅	reduce_mean/large_f32_threads=1-internal/262144	1.11 ms	1.20 ms	+8.3%
✅	matmul/large_generic_f32_threads=1/32x1024x1024	10.62 ms	11.19 ms	+5.4%
✅	matmul/large_generic_bf16_threads=8/32x1024x1024	2.16 ms	2.23 ms	+3.3%
✅	gather/medium_bf16_threads=1-internal/32768	2.81 µs	2.91 µs	+3.2%
✅	matmul/small_generic_f32_threads=8/1x256x256	52.89 µs	54.34 µs	+2.8%
✅	matmul/small_generic_f16_threads=8/1x256x256	39.87 µs	40.72 µs	+2.1%
✅	matmul/medium_generic_bf16_threads=1/32x512x512	593.22 µs	605.17 µs	+2.0%
✅	matmul/medium_generic_f32_threads=8/32x512x512	1.62 ms	1.65 ms	+1.6%
✅	matmul/medium_generic_f32_threads=1/32x512x512	2.76 ms	2.77 ms	+0.4%
✅	matmul/medium_generic_f16_threads=1/32x512x512	41.35 µs	41.10 µs	-0.6%
✅	matmul/small_generic_bf16_threads=1/1x256x256	40.81 µs	40.53 µs	-0.7%
✅	sampling_latency/top_k_per_token	556.18 µs	548.73 µs	-1.3%
✅	sampling_latency/top_p_per_token	1.14 ms	1.13 ms	-1.3%
✅	sampling_latency/min_p_per_token	386.36 µs	379.83 µs	-1.7%
✅	kv_cache/alloc_dealloc_pages	40.92 µs	40.20 µs	-1.8%
✅	tokenization/decode_tokens_per_second	7.56 ms	7.43 ms	-1.8%
✅	matmul/medium_generic_f16_threads=8/32x512x512	49.18 µs	47.90 µs	-2.6%
✅	matmul/small_generic_f32_threads=1/1x256x256	47.65 µs	46.37 µs	-2.7%
✅	matmul/large_generic_f16_threads=1/32x1024x1024	94.26 µs	91.10 µs	-3.3%
✅	gather/large_f32_threads=1-internal/131072	40.70 µs	39.00 µs	-4.2%
✅	sampling_latency/greedy_per_token	3.77 µs	3.60 µs	-4.3%
✅	matmul/large_generic_f16_threads=8/32x1024x1024	113.04 µs	107.88 µs	-4.6%
✅	matmul/small_generic_f16_threads=1/1x256x256	42.12 µs	39.89 µs	-5.3%
✅	gather/medium_f16_threads=1-internal/32768	2.96 µs	2.78 µs	-5.9%
✅	reduce_mean/medium_f32_threads=1-internal/65536	297.55 µs	277.73 µs	-6.7%
✅	add/large_f32_threads=1-internal/4194304	52.34 ms	47.90 ms	-8.5%
✅	reduce_mean/small_f32_threads=1-internal/4096	19.59 µs	17.44 µs	-11.0%
✅	gather/medium_f32_threads=1-internal/32768	5.05 µs	4.48 µs	-11.3%
✅	add/large_f16_threads=1-internal/4194304	52.74 ms	46.65 ms	-11.6%
✅	add/medium_bf16_threads=1-internal/262144	3.48 ms	3.01 ms	-13.6%
✅	gather/small_f32_threads=1-internal/4096	891.3 ns	766.7 ns	-14.0%
✅	gather/small_bf16_threads=1-internal/4096	618.5 ns	529.4 ns	-14.4%
✅	add/medium_f32_threads=1-internal/262144	3.32 ms	2.82 ms	-14.9%
🟢	grammar_masking/llguidance_compute_mask/32	104.39 µs	88.35 µs	-15.4%
🟢	matmul/large_generic_bf16_threads=1/32x1024x1024	2.76 ms	2.31 ms	-16.5%
🟢	add/large_bf16_threads=1-internal/4194304	58.34 ms	47.35 ms	-18.8%
🟢	add/small_bf16_threads=1-internal/1024	17.84 µs	14.31 µs	-19.8%
🟢	add/medium_f16_threads=1-internal/262144	3.65 ms	2.89 ms	-20.9%
🟢	gather/small_f16_threads=1-internal/4096	689.1 ns	527.6 ns	-23.4%
🟢	add/small_f16_threads=1-internal/1024	18.62 µs	14.18 µs	-23.9%
🟢	add/small_f32_threads=1-internal/1024	350.3 ns	214.5 ns	-38.8%

Visual flags: ⚠️ ≥ 15% slower, 🔴 ≥ 30% slower — calibrated against measured runner noise (~27% worst-case on multi-threaded matmul)

Host info

CPU: Apple M1 (Virtual)
Cores: 3
OS: Darwin 25.5.0 arm64
Rust: rustc 1.97.1 (8bab26f4f 2026-07-14)
Load avg: { 3.73 3.98 6.06 }

What this cannot catch

• Regressions in code paths not covered by these benchmarks (e.g., end-to-end decode with a real model)
• Sub-threshold regressions that compound over multiple PRs
• Performance changes that only manifest under GPU execution
• Latency changes in the ORT integration path (these benchmarks exercise the native Rust kernels)