feat(workflow): add self-improvement lifecycle to the enhancement module

[DTALEX66]

Scope

Absorb the personal-agent auto-growth (self-improvement) pattern into the Workflow Assistance enhancement module in neutral, cross-machine form.

• 10th managed skill workflow-assistance-self-improvement: usage sidecar + active/stale/archived lifecycle + pin opt-out + archive-not-delete + backup-before-transition + provenance filter; client-neutral (no agent-runtime dependency).
• scripts/workflow/skill_lifecycle.py (stdlib-only): status/record/run/archive/restore/pin/unpin/backup on any skills root; managed set = created_by: agent only — repository-owned module skills are protected (verified: live status shows all deployed skills repo-owned).
• 8 lifecycle regression tests: provenance filter, counter bumps, stale→archive with backup, pin bypass, archive/restore round-trip, non-managed refusal, dry-run, atomic sidecar write.
• Boundary contract nine→ten + managed_surfaces; boundary + sync tests updated; module doc updated.

Persists across machines: skills + tool ship in the module repo (sync deploys anywhere); auto-grown knowledge is promoted back into codex-assets via PR — the repo is the durable store, never a machine-local sidecar.

Evidence

• Targeted: lifecycle 8/8, sync 20/20, boundary 3/3, scanner OK, governance PASS; full quality gate QUALITY_GATE_PASS (21 gates).
• Live overlay applied + verified: 10 skills, preserved_user_config_fields=[].
• No CURRENT_STATE digest change (changed paths not in canonical digest inputs). Candidate SHA e87571c, tree 2dc8e6d.