Canvas AI: Give AI more context about the libraries present in Canvas

The prompts that i used for testing are:-

1. Create a component that has a formatted text area and gives me an option to change the background colour using a list of colour options
2. Can you please create a CTA component named test1 that allows me to use images with a bit of text and to include a CTA to link to other pages and some background colour.

The output must contain the imports for cn and FormattedText.

As per discussion with @tim.plunkett, @NarendraR and @balintbrews for now we can go with the hardcoded approach but we need to figure out a better way to dynamically parse the docs and pass the required libraries as a context to the LLM.
Added the descriptions for each libraries as per the request. This is ready for review again.

I have created a follow-up #3549430: Canvas AI: Find a better approach of passing the libraries supported by canvas as a context to LLM.

So #3549558: AI may generate Tailwind classes only compatible with older Tailwind versions issue has made the need of this issue more prominent. According to our discussion with Jamie, Markus, Tim, Narendra and Aanand we agreed with not proceeding the hardcoded approach in this issue.

Here's what i suggest until we get a better and a consolidated approach for this, We can find a way of reading the files that are present in the
/canvas/docs/user/src/content/docs/code-components and create temp file/set of instructions(generate a guide for LLM) which can be passed as a context to the LLM and we can avoid the usage of the libraries that are not present in Canvas.

Temporary advantages of this approach are:-

1. This will temporary solve the problem of previews being broken.
2. This approach is also better than the current approach as it removes the duplication of the instructions and it will also get the latest changes in the documentation.
3. This might also be helpful in handling issues such as #3549558: AI may generate Tailwind classes only compatible with older Tailwind versions which might otherwise need separate instructions in the prompt thus increasing the instructions.

One thing that popped up in my mind while testing this is say if a user adds a library which is not currently supported by canvas system and it tries to use that library then it will break the component preview.

I tested this with addition of a new library:-

[
    "name" => "lucide-react",
    "type" => "UI / Icons",
    "description" => "A lightweight, clean, and consistent icon set that is highly tree-shakeable and customizable via props.",
    "code" => "```jsx
import { Bell, User } from 'lucide-react';

export default function NavBar() {
  return (
    <nav className=\"flex items-center justify-between p-4 bg-white shadow-sm\">
      <span className=\"font-bold text-lg\">MyApp</span>
      <div className=\"flex gap-4\">
        {/* Pass standard props like size, color, or strokeWidth */}
        <Bell size={24} className=\"text-gray-600 hover:text-blue-500\" />
        <User size={24} strokeWidth={1.5} />
      </div>
    </nav>
  );
}
```",
  ],

with all the required fields that the form for libraries requires and i tried creating a code component using the following prompt:-
Create a React component called DynamicIconBox. The goal of this component is to render different icons based on a string prop passed to it, using the lucide-react library..

It does create a code component correctly but the preview for that component is broken with an error:-Failed to resolve module specifier "lucide-react". Relative references must start with either "/", "./", or "../".Understand this error

I don't think this form should be editable for now or else if it's editable then we should ensure that the required packages are imported at correct place by the user.

Also once we have included the libraries in a form then we should remove the code related to libraries in the agent instructions for ex: we have included examples of code for tailwind and cva in the component agent.

This is a much better approach than the form one earlier as we are anyways not adding the support for any libraries using canvas ai until canvas supports the libraries.

Just a small nit for a better message else everything looks good.

Removing tags as there are test coverages and also code looks good to me so moving it to RTBC.