feat: completion-enforcement loop (fallback, proof, emit-model, auditor)

Close the loop so dispatched work actually completes on bl:

- on_no_result fallback in followup-sweeper (op + job forms via
  exec-constrained registry / job-dispatch), seeded on the three
  autonomy-pulse jobs; fallback_due() dedupes the gravity path
- gravity.py: add __main__ entry (loop-remediator.timer was a no-op),
  300s re-arm budget, fallback firing + stamp/skip logic
- harvester: proof-of-result followups (result_has_evidence),
  acted-variant NACK, emit-model tool-hint wording
- envelope: RESPONSE RULE states the emit model (agents EMIT
  directives verbatim; runtime executes; works from bare containers)
- completion-audit.py + systemd 15-min timer: per-family funnel,
  swarm drain, followup backlog; digest DM when degraded, 6h heartbeat
- tests/test_completion.py (29 tests), JOB-SPEC.md docs

Tests: 67/67 focused green (completion + tool_calls).

bin/prompt_envelope.py

@@ -17,7 +17,10 @@ BOX_API = "https://box.muse-dev.online/api/box"
 
 def _response_rule():
     return (
-        "\nRESPONSE RULE: Execute your steps using [TOOL ...] directives or background tmux commands."
+        "\nRESPONSE RULE: Act by EMITTING [TOOL ...] / [DM ...] directive lines verbatim in your reply"
+        " — you do not run them yourself. The Box runtime on bl executes each directive"
+        " (this works from containers with no box CLI or tmux socket) and posts the result back here."
+        " Background tmux commands work too when you have a shell."
         " When complete, conclude your output with the [RESULT ...] line so the harvester records it.\n"
     )