release: prepare ml4t-diagnostic for stable 0.1.0

[stefan-jansen]

Summary

• add the cross-sectional quantile_profile API with deterministic per-date bucketing and explicit pooled behavior
• remediate the adversarial beta-exit findings across statistical correctness, error handling, dependency management, portability, public APIs, and rendering of unavailable results
• make every public README, quickstart, user-guide workflow, and API example executable and release-blocking
• add Linux, macOS, and Windows validation for Python 3.12, 3.13, and 3.14, including dependency security audits and installed-wheel smoke tests
• remove optional integrations, examples, and statistical modes that could not be validated for a stable release

Why

This prepares ml4t-diagnostic to move from beta to a stable 0.1.0 release. The branch treats incomplete inference, silent fallbacks, unsupported platform combinations, stale examples, and ambiguous unavailable values as release blockers.

User impact

The supported runtime contract is CPython 3.12 through 3.14 on Linux, macOS, and Windows. Breaking beta API changes are documented in the public migration guidance. Corrado event-study inference and other integrations that could not be validated were removed rather than retained with misleading behavior.

The package version remains 0.1.0b25 in this PR. The stable version bump and tag should occur only after this PR passes the complete remote matrix and merges.

Validation

• pre-commit run --all-files
• uv run ty check
• uv run pytest tests/ -q -n auto --timeout 120: 5,353 passed, 21 skipped
• executable public documentation and import contracts: 36 passed
• uv run mkdocs build --strict
• uv build
• release artifact policy verification for wheel and sdist
• isolated installed-wheel import smoke test
• twine check dist/*
• final independent code review: pass, no findings