Add LLM 1.0.0 extension

[JNK234]

Adds the LLM extension (codeName llm, v1.0.0) — a unified multi-provider LLM extension for NetLogo agents (OpenAI, Anthropic/Claude, Gemini, Ollama, OpenRouter, Together AI) with per-agent memory, sync/async chat, prompt templates, and reasoning-model support. Built with sbt packageZip.

[TheBizzle]

Please update this line to use NetLogo version 7.0.4 and rebuild your extension. If nothing else, your extension will then cease to bring in a dependency on the Matomo .jar.

Please update the URL here to be under the NetLogo organization. Same for the homepage field in libraries.conf.

Also, you guys have a bunch of demo models in the repository for this extension. Please include them inside the .zip for this extension. The NetLogo Models Library can automatically find them inside the unzipped folder for this extension and present them to users of the extension.

You can close the pull request for the 7.1 branch. I will cherry-pick these changes over to the 7.1 branch after they get merged onto 7.0 (assuming your 7.1 PR isn't any different than this one).

[TheBizzle]

It might also be good to explain in the longDescription that you've added to libraries.conf, or in messages directly from the extension, that people need to take extra steps when setting up this extension—namely: producing a config file and putting it in the right place. Otherwise, the likely course of events for them is:

1. Find the extension in the Extension Manager
2. Install it
3. Open a demo model
4. Get an error that it's not set up correctly, without having clear steps for fixing that

[JNK234]

Thanks for the review! I've pushed an update addressing everything:

• NetLogo version → 7.0.4 and rebuilt the extension; the new zip no longer brings in the Matomo .jar.
• Homepage now points to the NetLogo org repo (https://github.com/NetLogo/Netlogo-LLM-Extension), both in build.sbt and the homepage field here.
• Demo models bundled in the zip — five demos (baba-is-ai, color-sharing, social-deduction-game, topology-tournament, provider-sensitivity), each with its model and prompt templates, under demos/ so the Models Library can auto-discover them. Config files are deliberately excluded so no API keys ship.
• longDescription now explains the required setup step (create a config file with a provider + API key and llm:load-config it before use), so installers know to expect it.

I've also closed the 7.1 PR (#88) as you suggested — it's identical to this one, so feel free to cherry-pick after this merges.