perf: reuse GCC-PHAT TDOA; skip 16 kHz AudioNormalizer

Delay-and-Sum still uses the tutorial STFT/cov/GCC-PHAT/DelaySum/ISTFT
path. TDOA is refreshed every BEAMFORM_TDOA_EVERY hops (default 5).

config.py

@@ -41,6 +41,7 @@ class Config:
     # SpeechBrain Delay-and-Sum (GCC-PHAT) on C920 stereo before enhance
     beamform_mode: str = "auto"  # auto | force | off
     torch_num_threads: int = 1
+    beamform_tdoa_every: int = 5
     # misc
     publish_timeout_s: float = 60.0
     log_level: str = "INFO"
@@ -88,6 +89,7 @@ def load_config(env_path: Optional[Path] = None) -> Config:
         enhance_hop_s=float(os.environ.get("ENHANCE_HOP_S", "0.2")),
         beamform_mode=os.environ.get("BEAMFORM_MODE", "auto").lower(),
         torch_num_threads=int(os.environ.get("TORCH_NUM_THREADS", "1")),
+        beamform_tdoa_every=int(os.environ.get("BEAMFORM_TDOA_EVERY", "5")),
         publish_timeout_s=float(os.environ.get("PUBLISH_TIMEOUT_S", "60")),
         log_level=os.environ.get("LOG_LEVEL", "INFO").upper(),
         cameras=_parse_cameras(os.environ.get("CAMERAS", "all")),
@@ -100,4 +102,6 @@ def load_config(env_path: Optional[Path] = None) -> Config:
             f"BEAMFORM_MODE must be auto|force|off, got {cfg.beamform_mode}")
     if cfg.torch_num_threads < 1:
         raise ValueError("TORCH_NUM_THREADS must be >= 1")
+    if cfg.beamform_tdoa_every < 1:
+        raise ValueError("BEAMFORM_TDOA_EVERY must be >= 1")
     return cfg