feat(rewrite): add conditional combined workflow

[andreatnvidia]

Related Issue

Closes #237

Plan Document

plans/237/combined-rewrite-graph.md

Summary

Add an opt-in rewrite workflow that executes replacement-map generation, the initial rewrite, evaluation, and a bounded number of conditional repair iterations in one Data Designer graph.

The graph statically unrolls max_repair_iterations and uses SkipConfig.when to skip repair and downstream re-evaluation after each row passes. It supports zero or multiple repair rounds, preserves no-entity passthrough, row order, final metrics, repair counts, and review flags, and continues to execute through NddAdapter.run_workflow().

The legacy workflow remains the default. Data Designer does not yet expose terminal failed-column and seed-row provenance through its result API, so a dropped row in the combined graph can only be attributed to rewrite-combined. Keeping this path opt-in avoids reducing default failure observability while that upstream capability is developed.

For entity-bearing rewrite runs, the combined graph reduces the full pipeline's base Data Designer workflow count from five to three. The two detection workflows and the separate non-critical judge path are unchanged.

Benchmark evidence

The authoritative paired suite combined-rewrite-api-paired-20260807-r3 completed 30 pairs and 60 executions without failures. Median Data Designer workflow count decreased from five to three. Paired wall time was 1.81% slower and rewrite time was 1.58% slower, while utility and leakage were comparable.

The latency comparison was confounded by more combined-path repairs, 30,379 additional output tokens, and fixed variant ordering. It therefore establishes neither a speedup nor an intrinsic slowdown. Performance remains a regression guardrail rather than the merge rationale.

Remaining follow-up gates

• Add structured terminal failure provenance to Data Designer before making the combined graph the default.
• Compare peak memory, artifact size, and tail latency on mostly-skipped and mostly-repaired remote workloads.
• If latency is used as rollout evidence, rerun the paired benchmark with balanced ordering and equivalent repair decisions.
• After failure attribution is available, remove the opt-in flag and legacy orchestration in a follow-up.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactoring
• CI, release, or contributor workflow update

Contributor Checklist

• PR title follows Conventional Commits, for example fix: handle empty entity list
• Related issue is linked, or a maintainer-owned no-issue reason is documented above
• For non-trivial changes, a plan document is linked above, or the no-plan reason is documented above
• Public API impact checked; skills/anonymizer/SKILL.md updated if needed
• No real PII added to tests, docs, notebooks, fixtures, or artifacts
• No API keys, service tokens, private keys, credentials, or real endpoint secrets added

The bundled skill remains on the unchanged legacy default and does not need an update for this opt-in path.

Validation

• Commands run:
  • .venv/bin/ruff check --fix .
  • .venv/bin/ruff format .
  • TMPDIR=<short-temp-path> .venv/bin/pytest --basetemp=<short-temp-path>/pytest - 1,130 passed
  • focused real Data Designer conditional scheduling with zero, one, two, and exhausted repairs
  • paired SLURM suite described above - 60 executions, 0 failures
• Skipped checks or known failures:
  • provider e2e tests were not rerun;
  • remote scale validation and a corrected paired benchmark remain follow-up gates.

Documentation and Artifacts

• Docs updated, or not needed
• If docs changed: make docs-build passes locally
• If tutorial sources changed: notebooks regenerated with make convert-notebooks
• If e2e, benchmark, or model-provider behavior changed: relevant validation is listed above

The internal plan documents the rollout decision, benchmark interpretation, completed compatibility coverage, and remaining upstream and scale gates. No user-facing documentation or tutorial changed.