Reduce Ruff ignore list and tighten lint scope

[Anselmoo]

Problem

Ruff is configured with a very large ignore list and many per-file ignores, masking real issues and making it hard to rely on linting for quality signals. This technical debt blocks enforcing the new COCO/BBOB docstring work and doctest PR (#47).

Current ignores include complexity (C901), branching (PLR0912/15), arguments (PLR0913), naming (N803/N806), magic values (PLR2004), missing docs (D103/D107), numpy legacy RNG (NPY002), print statements (T201), ambiguous names (E741), long lines (E501), trailing commas (COM812), unused variables (F841), and broad per-file exceptions across tests, visualization, benchmarks, and core files.

Goal

Shrink the ignore surface to a minimal, intentional set; document what remains; and enable category-specific enforcement so new work (docstrings, doctests, COCO/BBOB) is not blocked by suppressed rules.

Proposal

1. Audit & classify rules

   • Identify which ignores are truly needed vs. legacy convenience
   • Group by theme: complexity/branching, naming, style, documentation, numpy RNG, prints, magic values

2. Reduce global ignores

   • Remove stylistic ignores that we can fix quickly (e.g., COM812, RET504, PERF401, SIM109/110 where trivial)
   • Re-enable documentation rules (D103, D107) once templates land (Create COCO/BBOB Docstring Template #65/Update Abstract Base Classes with COCO/BBOB Requirements #66/CI/Documentation Quality: Pre-commit Failures & Ruff Exception Reduction #68)
   • Re-enable NPY002 gradually with a clear plan (or document why retained)
   • Re-enable naming rules (N803/N806) in new/modified files first

3. Scope per-file ignores

   • Trim per-file ignores to only what’s justified; add comments explaining each
   • Keep test/visualization allowances but narrow lists (remove unused ones)

4. Introduce phased enforcement

   • Phase 0: Baseline report only (no fails) to map violations
   • Phase 1: Enforce on new/modified files (pre-commit target paths)
   • Phase 2: Enforce globally after cleanup

5. Add documentation

   • Document remaining ignores in pyproject.toml with rationale and sunset plan
   • Add a short docs/development/linting.md describing lint expectations

6. CI guardrail

   • Add a lint job that fails on newly introduced violations for rules we re-enable

Acceptance Criteria

• Global ignore list reduced to a minimal, documented set
• Per-file ignores trimmed and commented
• Report of current violations attached in the issue (Phase 0)
• Phase 1 policy: re-enabled rules enforced on new/modified files
• CI lint job updated to fail on re-enabled rules
• NPY002 and PLR2004 have a tracked remediation plan (or documented justification)
• Docstring rules (D103/D107) aligned with COCO/BBOB docstring rollout

Complexity

Medium — requires audit + targeted cleanups across categories.

Dependencies / Related

• PR Add comprehensive doctest support for all optimizer classes #47 (doctests) should benefit from stricter linting
• COCO/BBOB docstring work (Create COCO/BBOB Docstring Template #65-Document ECharts Architecture Decision and Fix SSR Issues #81) will re-enable doc rules

Suggested Next Steps

• Run uv run ruff check opt/ --output-format=full and attach summary
• Draft a proposed new ignore list and per-file ignores with comments
• Implement Phase 1 (new/modified files) in CI