feat: curated emoji denylist, vocab audio URLs in CSV

- Expanded _EMOJI_STOP from ~20 to ~80 keywords after manual review
  of all 2,261 emoji-word pairs. Removes false positives from
  polysemous words (french→🍟, water→🤽, rock→🪨, etc.)
- Emoji count: 2,261 → 1,820 (removed ~440 bad matches)
- hebrew_dict.csv now populated with audio_url from pealim.com scrape
  (8,727 words with audio URLs)
- Cached emoji_lookup.json (1,749 keywords from Unicode emoji-test.txt)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

apkg_builder.py

@@ -392,11 +392,35 @@ def _conj_audio_tag(slug: str, form_key: str) -> str:
     return ""
 
 
-# Stop-words excluded when matching emoji by meaning keyword
+# Keywords excluded when building emoji lookup AND matching meaning text.
+# Curated from manual review of all 2,261 emoji-word pairs (Sprint 8).
 _EMOJI_STOP = frozenset({
+    # Basic stop words
     "to", "be", "a", "an", "the", "of", "in", "on", "at", "for", "and",
-    "with", "by", "or", "but", "not", "as", "its", "face", "hand", "sign",
-    "symbol", "button", "small", "large", "light", "dark", "open", "closed",
+    "with", "by", "or", "but", "not", "as", "its",
+    # Generic emoji description words (too vague)
+    "face", "hand", "sign", "symbol", "button", "small", "large",
+    "light", "dark", "open", "closed",
+    # Numbers → clock emoji (🕐🕑🕒 etc.)
+    "one", "two", "three", "four", "five", "six", "seven", "eight",
+    "nine", "ten", "hundred", "thousand",
+    # UI/media buttons (⏭️▶️⏪⏯️🔁🔚🔜⏺️)
+    "next", "fast", "play", "pause", "repeat", "end", "soon", "record",
+    # Abstract words → misleading object emoji
+    "part", "place", "mark", "post", "department", "store", "note",
+    "control", "level", "stop", "cover", "roll", "rolling", "pick",
+    "over", "right", "way", "skin", "drop", "middle", "piece", "section",
+    # Country/direction words → flag emoji (🇰🇵🇬🇸🇮🇴🇻🇦🇨🇫 etc.)
+    "north", "south", "northern", "southern", "western", "eastern",
+    "central", "territory", "kingdom", "united", "virgin",
+    # Common words producing bad emoji matches
+    "new", "big", "full", "last", "first", "double", "slightly",
+    "without", "from", "behind", "people", "position", "status",
+    "situation", "game", "call", "trade", "male", "female", "person",
+    "letter",
+    # Polysemous words → wrong emoji sense
+    "french", "fried", "board", "bow", "water", "union", "rock",
+    "left", "back", "crane", "dash", "bar", "wheel", "horizontal",
 })