[DO NOT MERGE YET] wip: CLIP Vision becomes its own thing

[blessedcoolant]

Summary

CLIP Vision is currently a sub part of IP Adapter. There are other uses for CLIP Vision. Better to have it extracted now rather than later. This PR will try to accomplish that.

DONE

• CLIP Vision Model Loader Node - Used for loading the available CLIP Vision Models

• IP Adapter Node updated to use a plugged in CLIP Vision rather than a drop down
• CLIP Vision Models now display in the Model Manager

• CLIP Vision Models added to Starter Models

• Linear UI works now with the new CLIP Vision setup. (But the CLIP Vision Models need be manually set each model) --- We'll most likely make this a user config that they can set in the model editor.
• New Field For CLIP Vision
• [MAJOR] Implement CLIP Vision Safetensors support

TODO

• Update models.invoke.com to actually use the right CLIP Vision names.