Update playground.py

Add Google Imagen 4 to the playground model list

Retitle playground to Proxy Playground and name CLIP in the intro

Keep only the L* lightness span for skin tone

Replace L*/H* scatter with a darkest-to-lightest skin strip

Blank race labels until analysis; replace refusing VLM with CLIP

Make playground gallery thumbnails a uniform square size

Add on-page generation loader with live time estimate

Show generated images immediately, before race/skin analysis

Retry Replicate rate limits so 10-image playground runs can finish

Add proxy playground app

initial commit