Update Jan 27

-Add new default depth model implementing 'Depth Anything"
Should be better than base Midas was at the same speed while also supporting AMD cards hopefully.
The model mode was reset to auto so you'll have to reset it if you still want to use the ZoeDepth model.