Migrate seeded package to official NNS-python identity

[OVVO-Financial]

Motivation

• Finalize the repository identity migration from the seeded parity prototype into the official NNS Python package (seeded from the parity-hardened pyNNS-core-backed main after PR PyPI release pipeline + official ovvo-nns package name #10).
• Provide a stable distribution identity and build/install behavior so the native extension installs as nns._nnscore and the published package metadata targets the official name.
• Preserve all existing parity evidence, cache fixtures, and original R-test adoption while shifting documentation and CI to the new identity.
• Support a safe environment-variable migration that prefers NNS_R_CACHE_ONLY while continuing to accept the legacy PYNNS_R_CACHE_ONLY toggles during the transition.

Description

• Renamed the Python package layout and public imports from pynns → nns by moving src/pynns → src/nns, updating __init__ exports and all internal imports and tests to use nns (public import package: nns).
• Updated packaging metadata in pyproject.toml to name = "NNS", version = "1.0.0a0", and to build/install from src/nns (wheel packages and sdist includes updated).
• Reworked native build and install paths: CMakeLists.txt and the nanobind binding source are pointed at src/nns/_nnscore_bindings.cpp and the extension is installed into nns so the runtime module is nns._nnscore; runtime discovery changed in src/nns/_native.py.
• Migrated CI, docs, examples, and scripts: .github/workflows/native-backend-ci.yml now imports nns._nnscore and uses NNS_R_CACHE_ONLY, tests/_r.py and scripts/regenerate_r_cache.py accept both NNS_* and PYNNS_* environment toggles, and README/docs were updated to reference the distribution NNS, import nns, native extension nns._nnscore, cache-backed CI behavior, and parity scope/caveats.
• Preserved parity evidence and fixtures by adding them to sdist include and leaving the original files in place, including docs/parity_results.md, docs/parity_status.md, docs/parity_plan.md, docs/original_tests_adoption.md, docs/plot_parity_policy.md, docs/native_original_src_coverage.md, tests/_r_cache.json, tests/fixtures/original_tests_expected.json, and original_tests/.

Testing

• Ran python -m pytest -q tests/invariants (installed editable in the local venv) and observed 314 passed.
• Ran cache-only parity NNS_R_CACHE_ONLY=1 python -m pytest -q tests/parity and observed 1778 passed, 11 skipped (the 11 skips are intentional live-R-only practical examples).
• Ran the original-tests subset NNS_R_CACHE_ONLY=1 python -m pytest -q tests/parity/test_original_* and observed 12 passed.
• Confirmed legacy migration support by running PYNNS_R_CACHE_ONLY=1 python -m pytest -q tests/parity which produced the same 1778 passed, 11 skipped result.
• Static checks passed: ruff check . and mypy both reported no issues.
• Packaging/build: python -m build could not be executed in this restricted environment due to a missing build module / network fetch limits, but an offline build via the repo build driver (uv build) succeeded and produced dist/nns-1.0.0a0.tar.gz and dist/nns-1.0.0a0-*.whl with the native nns._nnscore shared module installed into the nns/ package.

No PyPI publication, release tagging, or changes to the vendored NNS-core behavior are included in this PR.

---

Codex Task