Implement ActEval 1.0 roadmap

[aminemanai2003]

What changed

• implement the v0.4 inference layer with reproducible bootstrap intervals,
  objective-aware paired comparisons, and fixed-bin calibration intervals
• implement the v0.5 monitoring/reporting layer with segments, temporal
  validation, weighted PSI drift, HTML reports, table exports, and plot export
• stabilize the documented public API as 1.0.0 with single-source versioning,
  Semantic Versioning, production metadata, migration/security/citation files,
  and cached metric signature validation
• audit and rewrite the README as a professional Python-library landing page

Why

ActEval 0.3 provided complete point, distribution, and decision-aware metrics,
but it lacked sampling uncertainty, governed reporting, portfolio monitoring,
and a formal stable API. This release completes those roadmap layers while
preserving the documented 0.3 API and the separation between actuarial
objectives.

User impact

Existing 0.3 calls continue to work. Users gain confidence intervals, paired
candidate/reference evaluation, segment and temporal reports, score drift,
standalone HTML/CSV/JSON artifacts, and a documented compatibility policy.

Validation

• 133 tests pass with 91.78% branch coverage
• Ruff format/lint and strict mypy pass, including Python 3.14 type mode
• wheel and sdist build and pass twine check
• metadata, stable classifiers, Apache license, citation, documentation, and
  py.typed contents were inspected
• clean isolated wheel installation, dependency check, and all-new-API smoke
  tests pass
• README public names, documentation targets, badges, CI, and PyPI links pass