Add --fp16-intermediates to use fp16 for intermediate values between nodes (#12953)

This is an experimental WIP option that might not work in your workflow but
should lower memory usage if it does.

Currently only the VAE and the load image node will output in fp16 when
this option is turned on.

comfy/model_management.py

@@ -1050,6 +1050,12 @@ def intermediate_device():
     else:
         return torch.device("cpu")
 
+def intermediate_dtype():
+    if args.fp16_intermediates:
+        return torch.float16
+    else:
+        return torch.float32
+
 def vae_device():
     if args.cpu_vae:
         return torch.device("cpu")