perf(cache): cache available models list for 24 hours

feat(ui): add loading state to submit button and exclude tts models

fix(gemini): filter available models to flash variants only, excluding pro and image models

feat(gemini): implement dynamic model selection in chat interface

fix(gemini): upgrade model to gemini-2.5-flash and remove debug route

fix: update gemini model version and add sys import

fix(genai): handle API key errors gracefully

Add .env to .gitignore and update genai_service to load API key from environment

refactor: reorganized project by separating routes and services