Fetch full event detail: batch listing omits text_fr/attachments for events

Found by testing: unlike news items, the agenda @results batch listing
always returns text_fr and items (attachments) as null for events - only
description_fr (short teaser) was ever available from it. The site itself
works around this by firing a second request per event
(@results?UID=<uid>&fullobjects=1) when you open one; we now do the same,
with a per-UID cache since a recurring event shares one UID across all its
occurrences.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

silly_rss.py

@@ -157,6 +157,29 @@ def build_news_items() -> list[str]:
     return entries
 
 
+_event_detail_cache: dict[str, dict | None] = {}
+
+
+def fetch_event_detail(uid: str) -> dict | None:
+    """
+    Contrairement aux actualites, l'API renvoie text_fr/items (texte complet
+    et pieces jointes) a null dans la liste paginee des evenements - il faut
+    une requete par UID pour les recuperer (c'est ce que fait le site
+    lui-meme quand on ouvre un evenement). Un evenement recurrent partage un
+    seul UID sur toutes ses occurrences, d'ou le cache pour eviter de
+    refaire la requete pour chaque occurrence.
+    """
+    if uid in _event_detail_cache:
+        return _event_detail_cache[uid]
+    try:
+        data = fetch_json(f"{EVENTS_API}?UID={uid}&fullobjects=1")
+        detail = (data.get("items") or [None])[0]
+    except Exception:
+        detail = None
+    _event_detail_cache[uid] = detail
+    return detail
+
+
 def format_event_meta(item: dict) -> str:
     parts = []
     start = parse_dt(item.get("start"))
@@ -183,7 +206,8 @@ def build_event_items() -> list[str]:
         guid = item.get("UID") or event_id
         link = f"{SITE}/agenda/{slugify(title)}?u={guid}"
         meta = format_event_meta(item)
-        description = (f"<p>{html.escape(meta)}</p>" if meta else "") + full_body_html(item, title)
+        detail = fetch_event_detail(guid) or item
+        description = (f"<p>{html.escape(meta)}</p>" if meta else "") + full_body_html(detail, title)
         pub_dt = parse_dt(item.get("created") or item.get("start"))
         # Un evenement recurrent partage le meme UID sur toutes ses
         # occurrences : on suffixe le guid avec la date de debut pour