Fix leaderboard loading error when no cache exists

Instead of trying to run evaluation on page load (which can fail with
network errors), show a user-friendly message asking them to click
'Re-evaluate All' to run the evaluation manually.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

leaderboard.py

@@ -190,18 +190,31 @@ def calculate_leaderboard_score(fertility: float, compression: float, unk_ratio:
 def get_cached_leaderboard(progress=gr.Progress()) -> Tuple[str, str, str]:
     """
     Get leaderboard results from cache if available.
-    If no cache exists, runs evaluation and caches results.
+    If no cache exists, shows a message to run evaluation.
     Returns: (leaderboard_html, per_dataset_html, status_message)
     """
     cache = load_leaderboard_cache()
     if cache:
+        # Also include any new submissions that were added after the cache
         return (
             cache.get("leaderboard_html", ""),
             cache.get("per_dataset_html", ""),
             cache.get("status", "") + "\n\n📦 *Loaded from cache. Click 'Re-evaluate All' to refresh.*"
         )
-    # No cache exists, run evaluation (first time only)
-    return run_leaderboard_evaluation(progress)
+
+    # No cache exists - show message to run evaluation
+    no_data_html = """
+    <div style="text-align: center; padding: 40px; background: #22272e; border-radius: 12px; border: 1px solid #30363d;">
+        <p style="color: #8b949e; font-size: 16px; margin-bottom: 16px;">📊 No evaluation data available yet.</p>
+        <p style="color: #e6edf3; font-size: 14px;">Click <strong>"Re-evaluate All"</strong> button above to run the full evaluation.</p>
+        <p style="color: #8b949e; font-size: 12px; margin-top: 12px;">This will evaluate all tokenizers on all 8 Arabic datasets (~5-10 minutes).</p>
+    </div>
+    """
+    return (
+        no_data_html,
+        no_data_html,
+        "⚠️ **No cached results found.** Click 'Re-evaluate All' to run the evaluation."
+    )
 
 
 def run_leaderboard_evaluation(