fix: protein bars → bag suffix, ice cream → tub suffix

Protein Gourmet Bar and Billion Protein Peanut Bar were hitting
supplement_keywords and getting suffix ' tub jar bag'. DINO then matched
the ice cream tub (a tub shape) for the bar query, stealing the box
from the actual Zero Ice Cream product.

- Add bar detection (regex \bbar\b) before supplement_keywords → ' bag'
- Add ice cream / gelato detection before all others → ' tub'
- Add speculoos, biscoff, lotus, pistachio, tiramisu, cheesecake
to flavors list so ice cream variant names are stripped correctly

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

app.py

@@ -134,6 +134,8 @@ def _clean_product_name(name: str) -> str:
         "salt", "coconut", "almond", "walnut", "cashew",
         # Legume/grain ingredients used as snack flavour variants (e.g. "Chips – Lentils & Pea Protein")
         "lentil", "lentils", "pea", "chickpea", "bean", "soy",
+        # Ice cream flavour variants
+        "speculoos", "biscoff", "lotus", "pistachio", "tiramisu", "cheesecake",
     ]
     parts = re.split(r'\s*[-–]\s*', clean_name, maxsplit=1)
     if len(parts) == 2:
@@ -168,7 +170,18 @@ def _get_packaging_suffix(name: str) -> str:
     if any(kw in name_lower for kw in wearable_keywords):
         return ""
 
-    # Ready meals / frozen food — checked FIRST so "Tandoori Chicken with Lime Rice"
+    # Ice cream / frozen desserts — checked FIRST so "Zero Ice Cream" doesn't fall through
+    # to the default " bag tub jar" and potentially match a bar bag instead
+    if "ice cream" in name_lower or "gelato" in name_lower or "frozen yogurt" in name_lower:
+        return " tub"
+
+    # Protein/snack bars — always sold in sealed bags, never tubs.
+    # Checked BEFORE supplement_keywords to prevent "Protein Bar" → " tub jar bag"
+    # which causes DINO to match ice cream tubs for bar queries.
+    if re.search(r'\bbar\b', name_lower):
+        return " bag"
+
+    # Ready meals / frozen food — checked BEFORE grain/canned so "Tandoori Chicken with Lime Rice"
     # doesn't fall into grain_keywords ("rice") or canned_keywords ("chicken breast")
     meal_keywords = [
         "meal", "bowl", "stew", "curry", "soup", "lasagna", "lasagne",