feat(cli): deprecation follows a superseded Source's provenance, and forget retires it without --force

[jasonssdev]

Summary

Deliverable 2 of "The Quiet Engine" arc (#1268), implementing the OpenSpec change openspec/changes/retire-superseded-sources/: deprecation follows provenance, and a superseded Source retires cleanly.

• Effective deprecation (bug: editing a watched source writes -N copies of every derived concept, and supersedes leaves its sole-source concepts stable #1259): a concept whose entire provenance is a superseded Source is deprecated at read time. A pure provenance_orphans predicate is folded into deprecated_concept_ids, so list STATUS, answer and retrieval candidates all agree through one predicate. Nothing is written to the concept file; the export input stays edge-only (the engine never reads its own export).
• Clean retirement (bug: retiring a superseded Source dead-ends — supersedes warns "not seeded" and forget needs --force, leaving a dangling reference #1263): forget <old-source> --scope source no longer refuses a superseded Source. In the same confirmed forget it removes the superseding Source's supersedes edge and detaches the old Source from surviving concepts — its generated provenance entry and its generated ## Related bullet (shape pinned in okf.related_bullet, "source this was extracted from"). Anything else still blocks: a hand-edited bullet, or a supersedes edge from a non-Source. purge is unchanged.
• relate supersedes no longer prints "not a seeded relation type".
• ADR-0046 (Proposed): effective deprecation follows the provenance of a superseded Source. docs/cli.md updated.

Spec deviation (applied to the forget-command delta and design D5): the generated ## Related bullet is pinned to the single phrase the builder and attach write, so a differently worded bullet counts as hand-written and still blocks.

Related issue

Closes #1259
Closes #1263
Refs #1268

Type of change

• feat — new feature
• fix — bug fix
• docs — documentation only
• refactor — no behavior change
• test — tests only
• chore / ci — tooling, build, or CI
• Breaking change

How was this tested?

• Repro tests observed RED first: forget sources/v1 --scope source --auto exited 1 on the supersedes edge and the inbound link (bug: retiring a superseded Source dead-ends — supersedes warns "not seeded" and forget needs --force, leaving a dangling reference #1263) and now exits 0; a sole-source concept of a superseded Source stayed stable in deprecated_concept_ids, list and answer (bug: editing a watched source writes -N copies of every derived concept, and supersedes leaves its sole-source concepts stable #1259) and is now deprecated at read time with its bytes unchanged; lint reports no dangling reference after the forget.
• Mutations, each killing a named test: the non-empty provenance guard, the sources/ root filter, the live-Source subset test, the generated-bullet matcher, the Source-authored superseder check, the rescan guard, and purge taking the retire path.
• After rebasing onto current main: ruff check, ruff format --check, mypy . (533 files), evals/run_self_tests.py, and pytest --cov (9639 passed, 2 skipped, 96.47%) pass.

Checklist

• My commits follow Conventional Commits.
• I added or updated tests for the change.
• I updated docs where behavior, interfaces, or the knowledge model changed.
• Lint, format, type check, and tests pass locally (ruff, mypy, pytest).
• Output remains OKF-conformant and derived stores stay reconstructible from the bundle + sources.
• The change is consistent with the project's guiding principles (local-first, provenance, freshness, human-in-the-loop).