Stage 3.17: cellpy 2.0.0 release checklist (benchmark acceptance, gates, tag)

[jepegit]

Goal

The 2.0 release checklist: benchmark acceptance, final gates, tag and publish.

Why

The release plan set a hard acceptance bar — no metric slower than 1.x, and the flip
(bridge removal + parquet) is expected to win on load and summary; if it doesn't, that
is a signal to investigate before shipping, not after.

Plan: architecture-plan/cellpy2-release-and-branching-plan.md §4 and §6;
architecture plan §6 row 3.6.

Gates

• Benchmarks vs the GHA-captured v1.x baselines: no metric slower; load and summary
  show the expected bridge-removal win (tiered gate per golden test fails #476).
• Value-parity oracle green on all golden cells, with the exception list explicit
  and named (no silently widened tolerances).
• Full CI suite green — essential (linux / uv) and full (linux / uv).
• DEPRECATIONS.md complete; migration guide published.
• Dependency budget applied; cellpycore pinned to an exact released version.
• Docs build published.
• Every other Stage-3 sub-issue closed.

Then

• Tag v2.0.0, publish to PyPI, update the conda feedstock.
• Confirm the v1.x branch state and start the 12-month bugfix-only window
  (decision Stage 0.11: Decision register — the calls that gate Stage 1+ #438-6) from this date.

[jepegit]

Additional task:

• Check alignment with guiding documents in .issueflows/04-designs-and-guides. Update documents if needed

[jepegit]

Phase C (rc1) shipped from clean master after #671.

• Release: https://github.com/jepegit/cellpy/releases/tag/v2.0.0rc1
• PyPI publish: https://github.com/jepegit/cellpy/actions/runs/30091209188 (green)
• Install: pip install --pre cellpy==2.0.0rc1

Still open for soak → stable v2.0.0 + feedstock + v1.x 12-month window start.