OpenRouter support with Google models

It's already possible to set OpenRouter as an endpoint, but it's not possible to select the model.

Doing `ai config` and selecting Model throws an error:
```
✖ Cannot read properties of undefined (reading 'value')
at LD.changeValue (file:///home/empz/.bun/install/global/node_modules/@clack/core/dist/index.mjs:14:4846)
at new LD (file:///home/empz/.bun/install/global/node_modules/@clack/core/dist/index.mjs:14:4514)
at Module.ve (file:///home/empz/.bun/install/global/node_modules/@clack/prompts/dist/index.mjs:22:294)
at showConfigUI (file:///home/empz/.bun/install/global/node_modules/@builder.io/ai-shell/dist/cli.mjs:1470:29)
at process.processTicksAndRejections (node:internal/process/task_queues:105:5)
at async file:///home/empz/.bun/install/global/node_modules/@builder.io/ai-shell/dist/cli.mjs:1512:9

ai-shell v1.0.12
```

I then tried
```
ai config set MODEL=google/gemini-2.0-flash-001
```

And it seemed to work, but the responses seems incomplete.

Running `ai -s list all files in cwd in bash` returns:

```
┌ AI Shell
│
◇ Your script:

-l

•
│
◆ Run this script?
│ ● ✅ Yes (Lets go!)
│ ○ 📝 Edit
│ ○ 🔁 Revise
│ ○ 📋 Copy
│ ○ ❌ Cancel
```

Reproduce the failure with `ai config` and the `MODEL=google/gemini-2.0-flash-001` setting, then trace the `showConfigUI` entry point named in the stack trace and the model-selection flow. Done means OpenRouter Google models can be selected through the configuration UI and `ai -s` produces a complete shell command.

Written by the indexing model from the issue text.