fix: declare unique_constraint for audit pkey in *_and_log / log_changes (OPS-4567)

[palantir-valiot]

Summary

Root cause: upsert_and_log (and the other *_and_log / log_changes paths) call repo.insert(%Changelog{...}) without declaring unique_constraint/3 for the audit table's pkey. When the sequence for audit_log/audit_logs is rewound or a concurrent upsert produces an id collision, Postgres raises on the unique pkey and Ecto surfaces it as Ecto.ConstraintError (see audit_logs_pkey in the stack). The existing error-handling in log_changes already swallows {:error, reason} returns (logs and returns {:ok, reason}), so turning the violation into a changeset error makes the main mutation succeed instead of blowing up the request.

Fix: changelog_changeset/1 now attaches unique_constraint(:id, name: ...) for the common pkey names customers actually use (audit_log_pkey, audit_logs_pkey, ${table}_pkey). Also upgraded benchee/credo within ranges (deps freshness per workflow).

Files changed:

• lib/ecto_trail/ecto_trail.ex — add @default_audit_table, changelog_changeset now builds candidates from Application.get_env(:ecto_trail, :table_name, ...) and calls add_unique_constraints_for_pkey/2.
• test/unit/ecto_trail_test.exs — added "constraint error handling for audit_log pkey (OPS-4567)" integration test exercising upsert_and_log under forced pkey collision (sequence rewind).
• test/unit/ecto_trail_constraint_unit_test.exs — new pure (no-DB) unit test asserting the constraint declarations are present on the internal builder.
• mix.exs — 1.0.4.
• CHANGELOG.md — concise entry + deps freshness note.
• mix.lock — benchee 1.5.1, credo 1.7.19 (and transitive).

Why

Linear: OPS-4567 — "jobs-lamosa-gto-prod: Root cause is missing unique_constraint handling". Triage: NOTIFY+FIX, severity high, category code_bug, first-party package. Matches the exact stack in the issue (ecto_trail 1.0.3 lib/ecto_trail/ecto_trail.ex:435: EctoTrail.log_changes/5 → repo.insert inside upsert_and_log).

Test plan

• mix format --check-formatted (clean).
• mix compile (clean; only pre-existing redundant-clause warning on an unrelated helper).
• Pure unit constraint regression test passes (standalone ExUnit run that loads only the compiled beams, no Postgres):
  • test/unit/ecto_trail_constraint_unit_test.exs asserts audit_log_pkey and audit_logs_pkey appear in the changeset constraints.
• Added TDD-style integration test under describe "constraint error handling for audit_log pkey (OPS-4567)":
  • Seeds a colliding row + rewinds the sequence, then calls upsert_and_log; asserts the mutation succeeds ({:ok, Resource}) instead of raising Ecto.ConstraintError.
• mix test executed (DB-dependent tests cannot connect in this pod; the test helper's Ecto.Migrator.run fails with connection refused as expected; the deterministic checks above cover the change).
• Upgraded benchee/credo within ranges (mix hex.outdated → mix deps.update); lockfile updated.
• Version bump + CHANGELOG entry added before the PR.
• Self-review: git diff inspected; no debug prints, no empty diff, no scope creep; branch name follows the NON-NEGOTIABLE palantir/OPS-4567-... convention.

Closes OPS-4567

[linear-code]

OPS-4567

[acrogenesis]

Closing as a duplicate of #24 — all of these PRs fix the same bug: Ecto.ConstraintError on audit_logs_pkey in EctoTrail.log_changes/5. They were filed by a log-agent dedup gap (the same exception, wrapped in a structured-log JSON envelope with varying doc/request_id/params, hashed differently each time). That gap is now fixed in palantir (commit 38438d6) so this won't recur. Consolidating on #24.