refactor: split outliers.py into a subpackage (#197)

[iabaako]

Pull Request Summary 🚀

What does this PR do? 📝

Splits src/datasure/checks/outliers.py (2,864 lines) into a subpackage, src/datasure/checks/outliers/:

• models.py — enums, Pydantic settings/result models, TAB_NAME
• compute.py — pure outlier/constraint computation functions and chart builders (no Streamlit calls)
• settings_ui.py — just outliers_report_settings
• report_ui.py — outliers_report, its report-rendering helpers, and the full column-configuration CRUD group (add/update/delete/validate/format-error/settings-table), which turned out to be called only from outliers_report, not outliers_report_settings, despite settings-sounding names — same pattern found in the two prior splits
• __init__.py — left empty, matching every other package in this repo

Also eliminates two intra-file duplicates found while splitting:

• _validate_constraint_settings/_validate_outlier_settings shared an identical try/except/format/return shape → extracted a generic _validate_settings(settings, model_cls, format_error) helper
• _render_constraint_violations_table/_render_outlier_table shared a near-identical "show more columns" expander/multiselect/save block → extracted _render_display_columns_expander(...)

Tests are mirrored into tests/checks/outliers/{conftest,test_models,test_compute,test_settings_ui,test_report_ui}.py, replacing the 3,528-line tests/checks/test_outliers.py, including the one pre-existing skip (test_stack_empty_dataframe, "Empty DataFrame causes Rust panic in Polars/Streamlit caching") preserved verbatim.

This is the fourth and last of the four splits tracked by #197 (enumerator in #265, backchecks in #266, gpschecks in #267), stacked on top of refactor/197-split-gpschecks-module since that PR is still open.

Why is this change needed? 🤔

Per #197: outliers.py mixed statistical/constraint computation with two distinct Streamlit UI flows (settings vs. report) in one 2,864-line file — the last of the four modules flagged as hard to navigate and review.

How was this implemented? 🛠️

Pure mechanical reorganization — no logic changes, aside from the two documented deduplications above (verified to produce identical output via the existing test suite). Grepped the full src/ tree to confirm only output_view_template.py imports from this module externally, and updated it to import outliers_report directly from datasure.checks.outliers.report_ui. The generated, gitignored output_view_1.py was left untouched, per this repo's convention of never editing generated output views directly.

How to test or reproduce ? 🧪

1. uv run python -m pytest tests/checks/outliers --cov=datasure.checks.outliers --cov-report=term-missing -q — 228/228 tests pass (exact match to the pre-split count, including the preserved skip), coverage improved from 97.44% to 97.47%
2. uv run python -m pytest -q -m "not slow" — full suite passes, same 7 pre-existing skips in the same places
3. Confirm the Outliers & Constraints report page in the app still renders and its settings still save/load correctly

Screenshots (if applicable) 📷

N/A — no UI or behavior changes.

Checklist ✅

• I have run and tested my changes locally
• I have limit this PR to less than 1000 lines of code change (if not, explain why) — ~4,700 lines changed, but this is a pure file-move/reorganization plus two small, verified dedup extractions; splitting it into multiple PRs would fragment a single mechanical refactor for no review benefit
• I have updated/added tests to cover my changes (if applicable)
• I have updated/added requirements to cover my changes (if applicable)
• I have run linting and formatting on any code changes (if applicable)
• I have updated the documentation (README, etc.) accordingly
• I have reviewed and resolved any merge conflict

🤖 Generated with Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
6 New issues
0 Accepted issues

Measures
0 Security Hotspots
97.3% Coverage on New Code
1.6% Duplication on New Code

See analysis details on SonarQube Cloud