docs(agents): add an AI usage policy and align the agent contracts

[amirhs1]

Summary

• Adds a root AI-POLICY.md: the human-facing governance document stating how AI
  coding assistants are used here, that the maintainer is accountable for every
  merged line, and that no AI system is an author of the software or of a release.
• Keeps the split explicit: AI-POLICY.md says what is permitted and who answers
  for it; AGENTS.md and CLAUDE.md say how the work is done and now defer to
  the policy without restating it.
• Carries three rules into AGENTS.md because they bind agents at work — a
  citation must be confirmed to exist and to support its claim, an expected test
  value must be derived rather than adopted from a model, and research data must
  not reach a third-party service.
• Extends the AGENTS.md canonical-sources list, which also omitted
  CHANGELOG.md, CITATION.cff, and docs/NAMING-CONVENTION.md.
• Adds a short ## AI assistance section to the README and a changelog entry.

Closes #24

Checks

• flake8 polyphys
• mypy polyphys/analyze polyphys/manage
• pytest README.md --doctest-glob="README.md" — 1 passed
• python -m build if packaging metadata, package data, or package layout changed
• python -m sphinx -W -b html docs/source docs/_build if docs/source/ changed

Documentation only: no package code, tests, or packaging metadata are touched, so
the linting, typing, build, and Sphinx checks have nothing to act on. The README
doctests were run because the README changed; they pass.

Scientific Correctness

• No measurement/statistics behavior changed
• Units, numerical fixtures, and domain assumptions are preserved or explained
• Parser lineage and organizer vocabulary are unchanged, or matching parser tests/docs were updated

Notes

Deliberately out of scope, and worth a separate decision:

• CONTRIBUTING.md does not exist, so contributor expectations live in
  section 8 of the policy for now. If a contributor guide is added later, that
  section should move and the policy should link to it.
• An AI-disclosure checkbox in .github/PULL_REQUEST_TEMPLATE.md would
  enforce section 8 at the point of contribution. Left out to keep this PR to the
  authorized scope.
• Zenodo / release metadata. Whether an AI-usage statement belongs in the
  archived record is a release decision, not a documentation one.
• The Sphinx site does not carry the policy. Adding docs/source/ai-policy.md
  would mean keeping two copies in sync; the README link is the single entry point
  for now.

Points to review before marking ready:

• Section 3 names Claude Code and OpenAI Codex as the assistants in use. Correct
  that list if it is incomplete.
• Section 2's authorship claim is the load-bearing sentence for a JOSS reviewer
  ([release] Evaluate MDAKit packaging and a JOSS submission #22) and for anyone citing the package; it should say exactly what you want on
  the record.
• Section 6 forbids sending unpublished simulation data and draft manuscripts to
  any AI service. That is stricter than current practice may be — confirm you want
  it stated as a rule.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.12%. Comparing base (47114a4) to head (fcfc10d).
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #25   +/-   ##
=======================================
  Coverage   85.12%   85.12%           
=======================================
  Files          14       14           
  Lines        1795     1795           
=======================================
  Hits         1528     1528           
  Misses        267      267           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.