feat(S2-02b): Implement AssemblyAI external transcription with speaker diarization

- Add assembly.ts module for REST API transcription via AssemblyAI
- Implement 5-step pipeline: upload → create job → poll status → download → save
- Enable speaker_labels for diarization (Speaker A, B, C...)
- Add millisecond-precision timestamps for each utterance
- Store JSON transcripts in storage/transcripts/{session_id}.json
- Add axios, dotenv dependencies
- Add transcribeLatest.ts helper for quick testing

User Story: S2-02b - Externe Transkription per REST API

package.json

@@ -0,0 +1,8 @@
+{
+  "dependencies": {
+    "dotenv": "^17.2.3"
+  },
+  "devDependencies": {
+    "@types/node": "^24.10.0"
+  }
+}