fix: add defaultModel and smallModel to kilo.jsonc

- defaultModel: qwen/qwen3.6-plus:free (main model for conversations)
- smallModel: openai/llama-3.1-8b-instant (for quick tasks)
- Configure models for built-in agents (code, ask, plan, debug)

This fixes Settings showing undefined models.

.kilo/kilo.jsonc

@@ -4,6 +4,8 @@
   "skills": {
     "paths": [".kilo/skills"]
   },
+  "defaultModel": "qwen/qwen3.6-plus:free",
+  "smallModel": "openai/llama-3.1-8b-instant",
   "agent": {
     "pipeline-runner": {
       "description": "Runs agent pipeline with Gitea logging",