W2.3 Bucket E: five GateStep CiUpsertStep rows (full ci_pipeline shadow bijection)

[briansrls]

Summary

W2.3 Bucket E (§2.3): add CiUpsertStep rows for all five GateSteps in ci_pipeline_step_ids_shadow, each with UpstreamUpsert inputs referencing the landed job shadow step.

• Five gate rows: structural_v2_compile, lens_ci_registry_signal, m1_rust_emit_probe_signal, testclaim_corpus_eval_signal, phase1_nat_semiring_rung_gate_signal.
• Helpers: ci_upsert_upstream_input, ci_upsert_upstream_job_input; gate verify/create via ci_gate_projection_node.
• Bijection witness now uses ci_upsert_steps_full_in_scope_step_ids (A+B+C+E) with full shadow-universe equality; gate shadow matching added alongside job rows.
• Historical A/B/C partial-slice ledger data retained; ci_upsert_steps_full_out_of_scope_step_ids is empty.

Completes W2.3 mechanical bijection for the current ci_pipeline step universe on main.

Test plan

• Forbidden grep §3 clean on src/v4/workflow/ci.dag
• CTRL_BUILD_WRAP_CARGO=0 cargo test -p v3-compiler v4_workflow_ci_runner_dag_smoke_test — 10/10 passed

Worker attestation

• Title describes the change
• PR body summarises what and why
• Tests run and named above
• No surprise commits (single Bucket E slice on ci.dag)
• No secrets / credentials / large binaries

[briansrls]

CI investigation (8212e77 → 5b9eadb): ci_v4 failed at v4 bootstrap compile, not in the Bucket E gate rows. Root cause: pre-existing #4063 rung_6.dag used data … = newline-before-call syntax; v4 parser rejects bare = at EOL (expected expression, found Newline @ L46). Same failure hit M1 emit probe and bootstrap gate result.

Fix (5b9eadb): Rewrote all 12 data rows in rung_6.dag to start the RHS call on the same line as =, matching rung_0_to_2_three_targets.dag continuation style. Bucket E ci.dag diff unchanged in substance.

— sent from loyal-badger-289