Campaign merge: distinct test_ids + per-test windowing (#507)

[jepegit]

Closes #507 — v2 Phases 1–2 (epic #402, themes V2-03/V2-07). Implements the
approved plan.

Campaign merge (V2-07)

CellpyCell.merge (previously dead code — wrong-arity caller, shadowed test) is
rewritten:

c1.merge(c2)                       # campaign (default): distinct tests
c1.merge(c2, mode="continuation")  # classic fold == from_raw([f1, f2])

Campaign semantics: distinct compact test_id per source stamped on raw
(overwrites tester-assigned ids — provenance stays in
meta_test_dependent.test_ID), per-test metadata records via #506's
Data.tests (id remap verified equivalent to core's merge_test_meta),
globally renumbered cycles (v1 — the legacy summary format can't represent
duplicates; renumber_cycles=False raises NotImplementedError, deferred to
#511), offset data points, unshifted timelines, no cumulative
carry-forward. Mixed cycle_modes store fine (warning); compute raises
MixedCycleModesError (#506 guard). Plus non-mutating merge_cells() and
test_meta.cycle_ranges_per_test().

Composite keys end-to-end (V2-03)

Raw test_id already survives into the core engine through the legacy bridge
and the step table groups on (test_id, cycle, step). The bridge strips the
column on the way out, so campaign objects (gated on _extra_tests) re-stamp
test_id onto the returned steps — the subsequent summary then flips
use_tid=True: joins on (test_id, cycle) and per-test windowing of
cumulative columns (they reset at test boundaries instead of running across
unrelated tests). Pinned by a real-data regression test that doubles as the
tripwire for future core bumps. Single-test and continuation objects are
untouched (safety pin asserts no test_id column appears).

Rode along

• Fixed latent _append merge_step_table bugs (gate flag only set inside the
  merge_summary block; offset read from t2.raw instead of t2.steps;
  last_cycle NameError when recalc=False).
• Replaced the dead/shadowed test_merge pair with a continuation-equivalence
  test (merge(mode="continuation") ≡ from_raw([f1, f2]) numerically).

Core's merge_data couldn't be used at the legacy seam (reads native schema
attribute names; verified AttributeError) — semantics mirrored; follow-up
core issue filed for bridge test_id preservation + legacy-schema support.

Verification

• New suite test_merge_campaign.py: 12 passed (incl. real-data per-test
  cumulative-reset pin and save/reload)
• uv run pytest -m essential: 99 passed
• Full uv run pytest: 627 passed, 62 skipped, 14 xfailed — goldens
  untouched

🤖 Generated with Claude Code

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[cursor]

Did not approve: Cursor Bugbot is not enabled on this account, so no automated review signal is available for this substantial reader/merge change. Assigned amundmr and juliawind for human review.

  

_{Sent by Cursor Approval Agent: Pull Request Router and Approver}