Better handle LLM generated PRs

It turns out LLMs are crap at following GitHub PR templates. What do people think about a light GitHub bot that spots that the PR template wasn't used and comments with a response that indicates they should switch to the PR template and if they're using AI, give them a prompt to make sure their submission is manageable?

I really don't enjoy reading 500 word PR descriptions for a +15 line change. It's possible to use LLMs in a concise manner, but it needs herding to get there. I think we can help contributors find that groove.