refactor: deprecate llama-omni-server, use llama-server built-in omni endpoints

modal/omni: PRODUCTION DEPLOY - text inference fully verified, llama-server running on T4

modal/omni: text inference verified (Chinese OK, English needs fix)

modal/omni: llama-server inference verified on T4

docs: add production script protect warnings to deploy guide

docs: comprehensive Modal deploy guide (both solutions + pitfalls)