Enforce rustfmt and clippy in CI

[tschm]

Closes #236.

rustfmt.toml declared a formatting standard for this crate, but no CI job enforced it, so the tree had drifted to 62 cargo fmt --check diffs across 9 files in src/algebra/dense/. Clippy was likewise never run in CI, and 4 warnings were shipping.

Changes

1. A lint job in ci.yml running cargo fmt --all -- --check and cargo clippy --all-targets -- -D warnings. It runs on macos-latest to match the existing build job, since sdp-accelerate is macOS-only.

2. The 4 warnings, fixed at source:

File	Lint	Fix
datamaps.rs:26	mismatched_lifetime_syntaxes	SparseExpansionCone<T> → SparseExpansionCone<'_, T>
supportedcone.rs:78	useless_borrows_in_formatting	dropped the redundant & in write!
reverse_compact.rs:48	unnecessary_unwrap	is_none()/unwrap() → if let Some((tree_index, clique_index))
augment_compact.rs:563	explicit_counter_loop	manual counter → .step_by(2).enumerate()

3. cargo fmt --all across the 9 drifted files.

One thing reviewers should know

cargo fmt --all alone does not make --check pass here. rustfmt reformats the tree but then exits 1 with error[internal]: left behind trailing whitespace on 6 lines inside #[rustfmt::skip] blocks — regions it refuses to rewrite yet still reports on. Those are stripped by hand in symv.rs (32, 33, 44, 45) and syr2k.rs (54, 55). Without that step the new lint job fails on a tree cargo fmt has just formatted.

Safety of the reflow

The src/algebra/dense/ changes are formatting only. Comparing each file against main with whitespace and commas stripped, 7 of 9 are byte-identical; the two that aren't (dense2x2/svd.rs, dense3x3/svd.rs) differ solely by rustfmt's default reorder_imports moving use crate::algebra::*;. No numeric literal is altered.

All four source fixes are MSRV-safe — '_ in path position has been stable since 1.31 — so the 1.70.0 leg of msrv.yml is unaffected.

Verification

cargo fmt --all -- --check                                    exit 0
cargo clippy --all-targets --features sdp-... -- -D warnings  exit 0
cargo clippy --all-targets -- -D warnings   (default feats)   exit 0
cargo test --features sdp-accelerate,faer-sparse,serde        235 passed, 0 failed, 0 ignored

Possible follow-ups (not in this PR)

• The lint job uses macOS runners to reach sdp-accelerate; ubuntu with sdp-netlib would be cheaper.
• Clippy covers the CI feature set but not python/julia, which would need a PyO3 toolchain in that job.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 94.73684% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 93.13%. Comparing base (b6aa042) to head (aa369cb).

Files with missing lines	Patch %	Lines
src/solver/core/cones/supportedcone.rs	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #238      +/-   ##
==========================================
+ Coverage   92.85%   93.13%   +0.28%     
==========================================
  Files          90       72      -18     
  Lines        7278     5743    -1535     
==========================================
- Hits         6758     5349    -1409     
+ Misses        520      394     -126     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[tschm]

Pushed a second commit (96f56dd) fixing an unrelated, pre-existing CI failure that was blocking this branch — and every other PR to this repo.

The MSRV job was failing before this branch existed (the last run on main also concluded failure):

error: package `serde_json v1.0.151` cannot be built because it requires
rustc 1.71 or newer, while the currently active rustc version is 1.70.0

Cargo.lock is gitignored, so each CI run resolves dependencies afresh and picks up releases that have raised their own MSRV past the rust-version = "1.70.0" declared in Cargo.toml. It is not only serde_json — I reproduced it against a real 1.70.0 toolchain locally and walked the chain: serde_json >=1.0.150, proc-macro2 >=1.0.107, unicode-ident >=1.0.23 and syn 3.x have all moved to 1.71. Pinning crate-by-crate would produce a long list that rots on the next release.

Instead the 1.70.0 leg now resolves with a current cargo under MSRV-aware resolution (CARGO_RESOLVER_INCOMPATIBLE_RUST_VERSIONS=fallback), which honours rust-version from Cargo.toml, then builds and tests with the pinned toolchain. One explicit pin remains and is commented in the workflow: serde requires serde_derive with an exact = version, and serde_derive 1.0.229 depends on syn 3.x, which the fallback resolver cannot see through.

Verified locally against an installed 1.70.0 toolchain: cargo build succeeds and cargo test passes 182/182 on default features. CI now confirms it — the 1.70.0, stable and beta legs all pass.

The stable and beta legs are untouched and still resolve freely.

Happy to split this into its own PR if you would rather keep this one to the lint change — it is a self-contained commit. The long-term fix is arguably to raise the declared MSRV to 1.71, but that is a semver-relevant policy call for maintainers, so I have not made it here.