fix: use h94/IP-Adapter repo for CLIP Vision safetensors

Changed CLIP Vision model source from openai/laion repos to h94/IP-Adapter
which provides the models in proper safetensors format that CLIPVisionLoader
can load directly.

Model sources:
- CLIP-ViT-H (SD 1.5): models/image_encoder/model.safetensors
- CLIP-ViT-bigG (SDXL): sdxl_models/image_encoder/model.safetensors

This fixes the "header too large" deserialization error caused by trying
to load PyTorch .bin files as safetensors.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

comfyui_models.yaml

@@ -208,7 +208,7 @@ model_categories:
   # SUPPORT MODELS (CLIP, IP-Adapter, etc.)
   # ==========================================================================
   support_models:
-    - repo_id: openai/clip-vit-large-patch14
+    - repo_id: h94/IP-Adapter
       description: CLIP H - For SD 1.5 IP-Adapter
       size_gb: 2
       essential: true
@@ -216,12 +216,12 @@ model_categories:
       type: clip_vision
       format: fp32
       vram_gb: 2
-      notes: Text-image understanding model
+      notes: Text-image understanding model from IP-Adapter repo
       files:
-        - source: "pytorch_model.bin"
+        - source: "models/image_encoder/model.safetensors"
           dest: "CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors"
 
-    - repo_id: laion/CLIP-ViT-bigG-14-laion2B-39B-b160k
+    - repo_id: h94/IP-Adapter
       description: CLIP G - For SDXL IP-Adapter
       size_gb: 7
       essential: true
@@ -229,9 +229,9 @@ model_categories:
       type: clip_vision
       format: fp32
       vram_gb: 4
-      notes: Larger CLIP model for SDXL
+      notes: Larger CLIP model for SDXL from IP-Adapter repo
       files:
-        - source: "open_clip_pytorch_model.bin"
+        - source: "sdxl_models/image_encoder/model.safetensors"
           dest: "CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors"
 
     - repo_id: google/siglip-so400m-patch14-384