feat(realism): LLM enrichment for user-class file bodies

Stage 6 of the realism migration. User-class file bodies (note,
todo, draft, script) optionally get LLM-authored content; system
classes (cron / daemon logs, /tmp caches) stay template-only because
formulaic *is* the right look for them.

New surface:

- realism.llm.circuit.LLMCircuitBreaker — process-local sliding-window
  breaker. 3 consecutive failures trip open; 60s cooldown to half-open;
  half-open success closes, failure re-opens. Protects the orchestrator
  tick from sustained Ollama wedges (per-call timeout already covers
  one-shot hangs).
- realism.prompts._style — em-dash suppression lifted from the
  email prompt. Persona.uses_llms_heavily opts out per the
  feedback_em_dash_llm_tell.md memory. Includes strip_em_dashes
  belt-and-braces sub for output that slipped past the prompt rule.
- realism.prompts.filebody — class-conditioned prompts (note / todo
  / draft / script) with persona context, language pinning, output
  shape rule.
- realism.bodies.make_body_with_llm — async wrapper around make_body
  that calls the LLM when one is provided AND the breaker allows.
  Falls back to template on timeout / error / empty / system-class.

Wiring:

- scheduler.pick_file accepts optional llm + llm_breaker + llm_timeout.
  When the planner picks a create action and the content_class is a
  user-class, the body_hint is replaced with the LLM-authored body
  (or falls back to the deterministic body_hint).
- orchestrator.worker constructs get_llm() at startup gated by
  DECNET_REALISM_LLM env var (any non-empty value enables; empty /
  "off" / "none" / "0" disables). Passes llm + breaker through every
  tick.
- decnet orchestrate gains --llm/--no-llm flag overriding the env var.

decnet/realism/prompts/_style.py

@@ -0,0 +1,39 @@
+"""Shared stylometric guards for LLM-bound prompts.
+
+Lifted from the original ``orchestrator.emailgen.prompt`` em-dash
+block so file-class prompts (note / todo / draft / script bodies)
+pick up the same suppression.  Per the
+``feedback_em_dash_llm_tell.md`` memory: em-dashes (—) are a strong
+LLM-authorship tell, suppress by default; allow only for personas
+explicitly opted in via ``EmailPersona.uses_llms_heavily``.
+"""
+from __future__ import annotations
+
+from decnet.realism.personas import EmailPersona
+
+
+_SUPPRESS_RULE = (
+    "Do NOT use em-dashes (—). Use commas, periods, or "
+    "parentheses instead. Em-dashes are a tell."
+)
+_ALLOW_RULE = (
+    "Em-dashes are fine — this persona uses them naturally. "
+    "Write in your usual style."
+)
+
+
+def em_dash_rule(persona: EmailPersona) -> str:
+    """Return the em-dash instruction line for *persona*'s prompt."""
+    if persona.uses_llms_heavily:
+        return _ALLOW_RULE
+    return _SUPPRESS_RULE
+
+
+def strip_em_dashes(text: str, persona: EmailPersona) -> str:
+    """Belt-and-braces: even with the prompt rule, small models leak
+    em-dashes occasionally.  Substitute with comma+space so the
+    output reads naturally; opt-in personas pass through unchanged.
+    """
+    if persona.uses_llms_heavily:
+        return text
+    return text.replace("—", ", ").replace("–", ", ")