Fix Raven GQA memory gate head expansion

Delete scripts/run_quasar_next_token.py

Add fixed FLA folder for Quasar generation

Remove Quasar FLA runtime patch from runner

Document direct Quasar generate cache fix

Use direct Quasar generate with FLA cache fix

Add simpler Quasar generation script

Add Raven dependency and Quasar next-token runner

initial commit