feat: add Prometheus metrics instrumentation

- New whisper_live/metrics.py with Counter, Gauge, Histogram metrics
- Track connections (opened/closed/rejected), transcription latency,
  audio processed, segments emitted, REST requests, and errors
- All metric helpers are no-ops when prometheus_client not installed
- --metrics_port CLI flag to expose /metrics endpoint (0 = disabled)
- Metrics integrated into server.py, base.py at key instrumentation points
- 17 new tests in tests/test_metrics.py (178 total passing)

run_server.py

@@ -90,6 +90,12 @@ if __name__ == "__main__":
         help='Expect raw PCM int16 audio from clients instead of float32. '
              'Audio will be normalized to float32 range [-1.0, 1.0].'
     )
+    parser.add_argument(
+        '--metrics_port',
+        type=int,
+        default=0,
+        help='Port for Prometheus /metrics endpoint. 0 = disabled (default). Requires prometheus_client.'
+    )
     args = parser.parse_args()
 
     if args.backend == "tensorrt":
@@ -120,4 +126,5 @@ if __name__ == "__main__":
         batch_max_size=args.batch_max_size,
         batch_window_ms=args.batch_window_ms,
         raw_pcm_input=args.raw_pcm_input,
+        metrics_port=args.metrics_port,
     )