Big re-write :)

I reordered the code in a more clean an organized way. Also this improves a lot
the emacs startup time again, bucause after some updates it becomes a bit slow
with the old configuration, so i did it again in a new way.

configs/init-ai.el

@@ -11,7 +11,6 @@
 
 ;; Cliente LLM (ollama, chatgpt, gemini, etc.)
 (use-package gptel
-  :defer t
   :config
   ;; (setq gptel-model 'gemma3:4b
   ;;       gptel-backend (gptel-make-ollama "Ollama"