feat: add apply and verify sync workflow

config.toml

@@ -33,3 +33,4 @@ target_lra = 11         # loudness range
 # duration = 30  # optional, seconds, null = full
 method = "phat"   # "phat" (GCC-PHAT, robust) or "standard" (raw)
 use_gpu = "auto"  # "auto" | "cpu" | "cuda" (ROCm/CUDA both via torch.cuda)
+out_dir = "output"  # synced output folder