Security review fixes: admin permissions, guest sessions, payment edge cases, assistant leaks

- Admin: refund/cancel/mark-resolved/Bazaar-link actions need change permission; question and
  curriculum import need add/delete; content browser needs view_question; guest device ids are
  shown to superusers only.
- Accounts: registering a guest's phone revokes the guest's earlier sessions; self-deletion keeps
  Redemption, UserOffer and the phone's OTP rows; the public demo guest cannot be deleted
  (DEMO_GUEST_DEVICE_IDS); display-name contact filter covers any separator and social handles.
- Payment: callback without Authority is 400; a second Bazaar purchase on a paid order is 409;
  orders snapshot credits/hearts/days at creation (payment.0011).
- Assistant: a broken diagram no longer makes a streamed answer free; a late off-topic marker
  refuses the answer (not stored, not cached); a disconnect saves the partial answer; only
  KNOWN_COLLECTIONS are searched.
- Infra: secret-key file is created atomically (workers no longer split keys); no default
  INTERNAL_SERVICE_KEY; Kavenegar key and Bazaar token kept out of error logs; Swagger/ReDoc
  pinned; schema throttled; docker compose bound to 127.0.0.1; non-ASCII digit ids are 400;
  ticket text capped at 5000.
- docs/security-before-production.md: remaining open items 24-31 and the guard tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

rag/services/generation.py

@@ -402,7 +402,15 @@ def stream_answer(question: str, context_chunks: list[dict] | None,
         if not clean:
             continue
         if released:
-            yield clean.replace(guardrails.OFF_TOPIC_MARKER, '')
+            if guardrails.OFF_TOPIC_MARKER in clean:
+                # علامت دیر آمد: بخشِ پیش از آن رفته است، ولی بقیه نمی‌رود و جواب «ردشده» حساب می‌شود
+                # تا نه ذخیره شود نه در کش برود (پیش‌تر کاربرِ بعدی همین متن را از کش می‌گرفت).
+                head = clean.split(guardrails.OFF_TOPIC_MARKER, 1)[0]
+                if head.strip():
+                    yield head
+                refused = True
+                continue
+            yield clean
             continue
         held += clean
         if guardrails.OFF_TOPIC_MARKER in held:
@@ -420,7 +428,7 @@ def stream_answer(question: str, context_chunks: list[dict] | None,
         result['diagram'] = None if refused else diagram_mod.fits(
             f'{question}\n{focus}\n{_previous_question(history)}', splitter.diagram)
     if refused:
-        yield guardrails.REFUSAL
+        yield ('\n\n' if released else '') + guardrails.REFUSAL
         return
     if not released and held.strip():
         yield held