feat: AI bug fixes, deck/collection migration, and enrichment pipeline

- Fix AI turn pipeline: add cmc to Scryfall cache, tag isRamp/isRemoval/
  isBoardWipe from oracle text during enrichment, enable post-combat spell
  casting for all difficulty levels
- Migrate deck builder, deck generator, and collection from ui/ to
  lan-client/ as iframe-embedded panels with LAB_API_BASE patching
- Fix ai-bridge.js: turnNumber, deck_name, battlefield serialization,
  5-playstyle mapping, window exports for cross-script access
- Fix engine.js: battlefieldCards flat array (not player.battlefield),
  aiDeclareAttacks extraction, combat overlay fixes
- Fix collection.js: template literal, column visibility alignment,
  set_code key, oracle text column, sortable category
- Add exile/graveyard last-card image rendering in ui.js
- Make state_encoder.py resilient to mixed input types
- Enrich type_line from Scryfall at card insert in deckbuilder.py

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

ml/encoder/state_encoder.py

@@ -141,8 +141,17 @@ class CardEmbeddingIndex:
             logger.error("Auto-download failed: %s", e)
             return None
 
-    def get_embedding(self, card_name: str) -> np.ndarray:
-        """Get embedding vector for a card name. Returns zero vector if not found."""
+    def get_embedding(self, card_name) -> np.ndarray:
+        """Get embedding vector for a card name. Returns zero vector if not found.
+
+        Accepts either a plain string or a dict with a 'name' key (defensive
+        against the LAN client sending structured battlefield objects).
+        """
+        # Handle dict objects from the LAN client's battlefield_detail
+        if isinstance(card_name, dict):
+            card_name = card_name.get('name', '')
+        if not card_name or not isinstance(card_name, str):
+            return self._zero_vec
         if not self._loaded:
             return self._zero_vec
         idx = self._name_to_idx.get(card_name.lower())
@@ -155,8 +164,11 @@ class CardEmbeddingIndex:
             return self._embeddings[idx]
         return self._zero_vec
 
-    def mean_pool_zone(self, card_names: List[str]) -> np.ndarray:
-        """Mean-pool embeddings for a list of card names (a game zone)."""
+    def mean_pool_zone(self, card_names: list) -> np.ndarray:
+        """Mean-pool embeddings for a list of card names (a game zone).
+
+        Each entry may be a plain string or a dict with a 'name' key.
+        """
         if not card_names:
             return self._zero_vec
         vectors = [self.get_embedding(name) for name in card_names]