feat: add VoicePrint tRPC router with service, ML engine, and storage modules

- Create voiceprint router with 7 protected procedures:
  getEnrollments, createEnrollment, deleteEnrollment, analyzeAudio,
  getAnalyses, getAnalysisResult, getJobStatus
- Create voiceprint service with core business logic:
  enrollment CRUD, audio analysis pipeline, alert creation, batch jobs
- Create ML engine with placeholder implementations for
  audio preprocessing, synthetic detection, voice matching, embedding
- Create storage module for audio file persistence on disk
- Add valibot schemas for input validation
- Wire voiceprint router into root tRPC router
- Add comprehensive unit tests (33 tests, all passing)

web/src/server/api/schemas/voiceprint.ts

@@ -0,0 +1,29 @@
+import { object, string, minLength, optional, number, picklist } from "valibot";
+
+export const CreateEnrollmentSchema = object({
+  name: string([minLength(1)]),
+  audioBase64: string([minLength(1)]),
+});
+
+export const DeleteEnrollmentSchema = object({
+  enrollmentId: string([minLength(1)]),
+});
+
+export const AnalyzeAudioSchema = object({
+  audioBase64: string([minLength(1)]),
+  enrollmentId: optional(string()),
+});
+
+export const AnalysisFilterSchema = object({
+  page: optional(number(), 1),
+  limit: optional(number(), 20),
+  verdict: optional(picklist(["NATURAL", "SYNTHETIC", "UNCERTAIN"])),
+});
+
+export const AnalysisResultSchema = object({
+  analysisId: string([minLength(1)]),
+});
+
+export const JobStatusSchema = object({
+  jobId: string([minLength(1)]),
+});