feat: dish recognition UX, background mode, and backend bug fixes

Flutter client:
- Progress dialog: redesigned with pulsing animated icon, info hint about
  background mode, full-width Minimize button; dismiss signal via ValueNotifier
  so the dialog always closes regardless of widget lifecycle
- Background recognition: when user taps Minimize, wasMinimizedByUser flag is
  set; on completion a snackbar is shown instead of opening DishResultSheet
  directly; snackbar action opens the sheet on demand
- Fix dialog spinning forever: finally block guarantees dismissSignal=true on
  all exit paths including early returns from context.mounted checks
- Fix DishResultSheet not appearing: add ValueKey to _DailyMealsSection and
  meal card Padding so Flutter reuses elements when _TodayJobsWidget is
  inserted/removed from the SliverChildListDelegate list
- todayJobsProvider refresh: added refresh() method; called after job submit
  and on DishJobDone; all ref.read() calls guarded with context.mounted checks
- food_search_sheet: scan buttons replaced with full-width stacked OutlinedButtons
- app.dart: WidgetsBindingObserver refreshes scan providers on app resume
- L10n: added dishRecognitionHint and minimize keys to all 12 locales

Backend:
- migrations/003: ALTER TYPE recipe_source ADD VALUE 'recommendation' to fix
  22P02 error in GET /home/summary -> getRecommendations()
- item_enricher: normalizeProductCategory() validates AI-returned category
  against known slugs, falls back to "other" — fixes products_category_fkey
  FK violation during receipt recognition
- recognition prompt: enumerate valid categories so AI returns correct values

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

backend/internal/adapters/openai/recognition.go

@@ -202,7 +202,8 @@ Return ONLY valid JSON without markdown:
   "fat_per_100g": 1,
   "carbs_per_100g": 0,
   "storage_days": 3
-}`, name)
+}
+"category" must be exactly one of: dairy, meat, produce, bakery, frozen, beverages, other`, name)
 
 	messages := []map[string]string{
 		{"role": "user", "content": prompt},

backend/internal/domain/recognition/item_enricher.go

@@ -3,11 +3,30 @@ package recognition
 import (
 	"context"
 	"log/slog"
+	"strings"
 
 	"github.com/food-ai/backend/internal/adapters/ai"
 	"github.com/food-ai/backend/internal/domain/product"
 )
 
+// validProductCategories mirrors the product_categories slugs seeded in the DB.
+var validProductCategories = map[string]struct{}{
+	"dairy": {}, "meat": {}, "produce": {}, "bakery": {},
+	"frozen": {}, "beverages": {}, "other": {},
+}
+
+// normalizeProductCategory returns a pointer to a valid product_categories slug.
+// It lowercases and trims the AI-returned value; if it is not recognised it falls
+// back to "other" rather than letting an invalid string reach the FK constraint.
+func normalizeProductCategory(category string) *string {
+	normalized := strings.ToLower(strings.TrimSpace(category))
+	if _, ok := validProductCategories[normalized]; ok {
+		return &normalized
+	}
+	fallback := "other"
+	return &fallback
+}
+
 // itemEnricher matches recognized items against the product catalog,
 // triggering AI classification for unknown items.
 // Extracted from Handler so both the HTTP handler and the product worker pool can use it.
@@ -80,7 +99,7 @@ func (enricher *itemEnricher) saveClassification(enrichContext context.Context,
 
 	catalogProduct := &product.Product{
 		CanonicalName:   classification.CanonicalName,
-		Category:        strPtr(classification.Category),
+		Category:        normalizeProductCategory(classification.Category),
 		DefaultUnit:     strPtr(classification.DefaultUnit),
 		CaloriesPer100g: classification.CaloriesPer100g,
 		ProteinPer100g:  classification.ProteinPer100g,