FRE-5006: VoicePrint quality improvements

- P2-1: Consolidate mock ML logic to Python canonical source
- P2-2: Fix weak hashes with SHA-256
- P2-3: Parallelize batch processing with Promise.allSettled()
- P2-4: Add DI pattern support to services
- P2-5: Add structured logging utility
- P3-2: Persist batch jobId for result retrieval

Co-Authored-By: Paperclip <noreply@paperclip.ing>

packages/db/prisma/schema.prisma

@@ -52,6 +52,25 @@ enum UserRole {
   support
 }
 
+enum DetectionVerdict {
+  NATURAL
+  SYNTHETIC
+  UNCERTAIN
+}
+
+enum AnalysisType {
+  SYNTHETIC_DETECTION
+  VOICE_MATCH
+  BATCH
+}
+
+enum AnalysisJobStatus {
+  PENDING
+  RUNNING
+  COMPLETED
+  FAILED
+}
+
 model Account {
   id                String    @id @default(uuid())
   userId            String
@@ -337,6 +356,44 @@ model VoiceAnalysis {
   @@index([audioHash])
 }
 
+model AnalysisJob {
+  id              String        @id @default(uuid())
+  userId          String
+  analysisType    AnalysisType
+  audioFilePath   String
+  status          AnalysisJobStatus
+  errorMessage    String?
+  completedAt     DateTime?
+  createdAt       DateTime      @default(now())
+  
+  user            User          @relation(fields: [userId], references: [id])
+  result          AnalysisResult?
+  
+  @@index([userId])
+  @@index([status])
+  @@index([createdAt])
+}
+
+model AnalysisResult {
+  id                  String    @id @default(uuid())
+  analysisJobId       String
+  syntheticScore      Float
+  verdict             DetectionVerdict
+  confidence          Float
+  processingTimeMs    Int
+  matchedEnrollmentId String?
+  matchedSimilarity   Float?
+  modelVersion        String?
+  
+  analysisJob         AnalysisJob @relation(fields: [analysisJobId], references: [id])
+  
+  createdAt           DateTime  @default(now())
+
+  @@index([analysisJobId])
+  @@index([syntheticScore])
+  @@index([verdict])
+}
+
 // ============================================
 // SpamShield Models (Spam Detection)
 // ============================================