feat: persist voice samples to disk, generate VTT, cleanup after analysis

test/testutil/fake_speech_transcriber.go

@@ -1,14 +1,18 @@
 package testutil
 
-import "context"
+import (
+	"context"
+
+	"github.com/paramah/gw_telegram/internal/domain/port"
+)
 
 type FakeSpeechTranscriber struct {
-	Transcript string
-	Error      error
-	CallCount  int
+	Result    port.TranscriptionResult
+	Error     error
+	CallCount int
 }
 
-func (f *FakeSpeechTranscriber) Transcribe(_ context.Context, _ []byte, _ string) (string, error) {
+func (f *FakeSpeechTranscriber) Transcribe(_ context.Context, _ []byte, _ string) (port.TranscriptionResult, error) {
 	f.CallCount++
-	return f.Transcript, f.Error
+	return f.Result, f.Error
 }