Add HALO Debug Pack - AMD ROCm bf16 diagnostic tools (#2562)

Adds diagnostic nodes for debugging black image issues caused by
numpy's lack of bfloat16 support on AMD ROCm, older NVIDIA cards,
Apple Silicon, and unified memory systems.

Nodes included:
- HALO VAE Decode/Encode (FP32) - forces fp32 to fix bf16 conversion
- HALO Latent Debug - check sampler output for NaN
- HALO Conditioning Debug - check text encoder for NaN
- HALO Model Debug - inspect model dtype

Repository: https://github.com/bkpaine1/halo_pack

Co-authored-by: bkpaine1 <bkpaine1@github.com>

custom-node-list.json

@@ -43306,6 +43306,17 @@
             ],
             "install_type": "unzip",
             "description": "This is a node to convert an image into a CMYK Halftone dot image."
+        },
+        {
+            "author": "bkpaine1",
+            "title": "HALO Debug Pack",
+            "id": "halo-pack",
+            "reference": "https://github.com/bkpaine1/halo_pack",
+            "files": [
+                "https://github.com/bkpaine1/halo_pack"
+            ],
+            "install_type": "git-clone",
+            "description": "AMD ROCm bf16 diagnostic tools for ComfyUI. Fixes black image issues caused by numpy's lack of bfloat16 support. Includes FP32 VAE decode/encode and debug nodes for latent, conditioning, and model inspection. Useful for AMD Strix Halo, older NVIDIA cards, Apple Silicon, and unified memory systems."
         }
     ]
 }