Resume retries after cooldown instead of giving up for the day

Previously 3 failed attempts at 06:00 (e.g. llamaswap model cold/starting)
permanently blocked generation until the next day. Now the retry loop
pauses SLOW_RETRY_MINUTES (default 60) after the burst budget is spent,
then resumes — one attempt per hour until jokes exist.

app/config.py

@@ -33,6 +33,9 @@ class Settings:
     retry_minutes: int = field(
         default_factory=lambda: int(os.environ.get("RETRY_MINUTES", "15"))
     )
+    slow_retry_minutes: int = field(
+        default_factory=lambda: int(os.environ.get("SLOW_RETRY_MINUTES", "60"))
+    )
     wikipedia_lang: str = field(
         default_factory=lambda: os.environ.get("WIKIPEDIA_LANG", "en")
     )