chore: make the workspace clippy-clean under -D warnings

[Sean-Koval]

Closes #34

Pure refactor: makes cargo clippy --workspace --all-targets --all-features -- -D warnings pass and makes CI, release and just clippy run that one command. No numeric behaviour, public Rust API, or Python-visible name/signature changes.

Warning counts

Counted as unique (lint, file, line, column) diagnostics from cargo clippy --workspace --all-targets --all-features --message-format=json (clippy 1.88.0), de-duplicated across the lib/test builds of the same file. (The issue's "~160 warning lines" counts the duplicated text output.)

	unique warnings
Before (origin/main @ d3e6dfa)	143
After clippy --fix	108
After this PR	0

Fixed vs. allowed per lint

lint	before	fixed	allowed
legacy_numeric_constants	35	35	0
type_complexity	29	29 (type aliases)	0
too_many_arguments	21	1 (params struct)	20
needless_range_loop	16	16	0
uninlined_format_args	8	8	0
manual_range_contains	6	6	0
manual_range_patterns	3	3	0
manual_clamp	2	2	0
iter_cloned_collect	2	2	0
useless_conversion	2	2	0
get_first	2	2	0
default_constructed_unit_structs	2	2	0
unnecessary_cast	2	2	0
useless_vec	2	2	0
unnecessary_to_owned	1	0	1
large_enum_variant, unnecessary_map_or, manual_memcpy, let_and_return, vec_init_then_push, needless_borrow, derivable_impls, len_zero, field_reassign_with_default, redundant_closure	1 each (10)	10	0
total	143	122	21

Every #[allow] added (21)

No crate/module-level allows; clippy.toml thresholds untouched.

#[allow(clippy::too_many_arguments)]:

item	reason
openquant::hyperparameter_tuning::randomized_search (8)	Public search API: grid_search's arguments plus n_iter and seed; signature kept stable
openquant::labeling::get_events (8)	Mirrors the mlfinlab get_events signature
pyopenquant::helpers::build_ohlcv_columns (8)	Takes the raw OHLCV columns it validates into OhlcvColumns; a params struct would duplicate that type
pyopenquant #[pyfunction]s: cla_allocate, data_clean_ohlcv, data_quality_report, data_align_calendar, hcaa_allocate, labeling_triple_barrier_events, labeling_triple_barrier_labels, labeling_meta_labels, labeling_get_events, pipeline_run_mid_frequency_pipeline, portfolio_allocate_with_solution, sb_fit_predict_classifier, sb_fit_predict_regressor, sr_estimate_strategy_failure_probability, sbt_generate_ou_paths, sbt_run_synthetic_otr_workflow, sbt_search_optimal_trading_rule (17)	Python keyword signature

#[allow(clippy::unnecessary_to_owned)]:

item	reason
MicrostructuralFeaturesGenerator::new_from_csv, tick_num_iter field init	The generator owns its thresholds (vec::IntoIter<usize>); clippy's borrowed-iterator suggestion would add a lifetime to a public type

Notable hand fixes

• needless_range_loop: rewritten with iter_mut().enumerate() / .take().skip(); iteration order and float accumulation order are unchanged. sampling::num_concurrent_events deliberately uses .take(end+1).skip(start) rather than a slice so start > end_idx stays a no-op instead of becoming a panic.
• manual_clamp in bet_sizing::discrete_signal: f64::clamp(-1.0, 1.0) matches the old if-chain, including NaN pass-through.
• large_enum_variant: private etf_trick::Source::InMemory now holds a Box<InMemoryTables>.
• too_many_arguments: private-to-the-crate helper pyopenquant::helpers::build_labeling_events takes a new LabelingEventArgs struct.
• New public aliases (transparent, no signature change): bet_sizing::{ReserveBetSizeRow, MixtureParams}, cross_validation::TrainTestSplit.

Test files

No expected value, tolerance or fixture changed, and no test was removed or ignored. However, --all-targets -D warnings also lints tests, so clippy --fix made semantically identical rewrites inside some test assertions, e.g. assert!(b >= -1.0 && b <= 1.0) -> assert!((-1.0..=1.0).contains(&b)), matches!(bin, -1 | 0 | 1) -> matches!(bin, -1..=1), pca[0].len() >= 1 -> !pca[0].is_empty(), inlined format args in assertion messages, vec![..] -> [..] for two literal lists (same values). Please review crates/openquant/tests/ with that in mind.

CI alignment

ci.yml, release.yml and justfile now all run exactly:

cargo clippy --workspace --all-targets --all-features -- -D warnings

The clippy.toml comment quotes that command (now true); docs/publishing.md and the docs-site local-build page were updated to match.

Caveat: workflows use dtolnay/rust-toolchain@stable (unpinned), and this was verified locally on clippy 1.88.0 only. A newer stable clippy may have additional lints; CI on this PR is the check for that.

Commands run (local, rustc/clippy 1.88.0)

command	result
cargo clippy --fix --workspace --all-targets --all-features --allow-dirty	143 -> 108 unique warnings
cargo clippy --workspace --all-targets --all-features -- -D warnings	Finished dev profile, exit 0
cargo test --workspace	exit 0; 40 test binaries/doc-test sets ok, 172 passed, 0 failed (the pre-existing #[ignore] on test_sadf_test is untouched)
cargo fmt --check	exit 0

Not run: Python bindings build (maturin) / Python tests, docs-site API-drift gate, benches (compiled under clippy only).

Observed but intentionally not changed

• MicrostructuralFeaturesGenerator::new_from_csv seeds current_bar_tick from tick_num_series.first() but the iterator is not advanced, so the first threshold appears to be consumed twice in get_features. Looks like a possible porting bug vs. mlfinlab's generator; behaviour left as is.

🤖 Generated with Claude Code

[Sean-Koval]

Review follow-up (pushed b003c10):

• CI's floating stable (1.98) flagged 6 sites that local 1.88 could not see. Fixed, all behavior-preserving: unnecessary_unwrap ×2 (cla, codependence — the match keeps a NaN correlation on the bivariate branch as before), manual_is_multiple_of (hpc_parallel; progress_every is validated > 0 so it is equivalent), a dead assignment to sell_volume (streaming_hpc), needless_range_loop in a bench, useless_conversion in a test.
• Added rust-toolchain.toml pinning 1.98.1 so -D warnings cannot break unrelated PRs when a new Rust ships.
• Verified locally on 1.98.1: clippy -D warnings exit 0 (also clean on 1.88), cargo fmt --check clean, cargo test --workspace 243 passed / 0 failed / 1 ignored.
• Test-file edits reviewed: all are equivalent rewrites; no expected value, tolerance or fixture changed.
• docs-checks fails here for a reason unrelated to this PR (depth-1 checkout breaks the freshness gate); the fix is in chore: adopt ai-dlc and record the production-readiness backlog #62. Update this branch after chore: adopt ai-dlc and record the production-readiness backlog #62 merges.
• Real bug found during review, deliberately not changed here: PurgedKFold::split leaks: test window starts at the first test label's END time #65 (PurgedKFold purge window).