Add interactive Quantum Katas learning experience to VS Code extension

[minestarks]

Previously, the Quantum Katas were only accessible through the website. This PR adds a full in-editor learning experience, including an AI tutor via Copilot Chat.

What's new

• Activity bar panel — A "Microsoft Quantum" sidebar with a tree view showing all Katas units and activities, with completion tracking (checkmarks, progress counts).
• Lesson panel — A webview that renders lesson text, runnable code examples, and exercises inline. Includes an action bar for navigation (Next/Back), Run, Check, Hint (via chat), and Explain (via chat).
• Workspace-based progress — On first use, a qdk-learning/ subfolder is scaffolded in the workspace with .qs files for each exercise. Progress is persisted in qdk-learning.json.
• Copilot agent and tools — A QDK Learning chat agent backed by ~12 language model tools (show, next, previous, goto, run, check, hint, solution, reset, read-code, get-state, list-units). The agent acts as a tutor: gives hints incrementally, explains concepts, and checks solutions.
• Code lens — "Check Solution" / "Reset Exercise" lenses appear on exercise files.
• Welcome view — Shown in the sidebar when no learning workspace is detected; prompts initialization.

Interesting files

• service.ts is the core singleton — manages catalog loading, navigation state, progress file I/O, exercise checking (via the existing runProgram infrastructure), and Q# file scaffolding.
• panel.ts manages the webview lifecycle; the webview app itself is a React/TSX app in src/learning/webview/.
• learningTools.ts wraps the service for Copilot tool invocations, including a confirmInit flow that prompts workspace creation on first use.

To try it

1. Build the extension (.build.py --wasm --npm --vscode)
2. Open a folder in VS Code, click the Quantum activity bar icon
3. Confirm workspace setup when prompted
4. Navigate lessons/exercises in the tree or panel, or chat with the @QDK Learning agent

description written with heavy help from Copilot

[amcasey]

The only thing I really care about is having copilot/eval know about the current editor contents - the rest of my comments are just hygiene. LGTM

[amcasey]

[joao-boechat]

nit: should "agents", "prompts", "skills", all be in a folder centralizing all ai-related files? Just to make sure that in the future it's easy to find all of the md files and modify their behavior if necessary.

[joao-boechat]

Some of copilot thinking mentions the tool calls and internal things that the user doesn't really need to see. Would be nice if there was a way to disable showing that, the experience would be more fluid

[joao-boechat]

The experience is very fluid and even better than the one we had at the website. I was skeptical at the idea of running everything behind copilot but somehow you made it very easy to use, can't wait to have it shipped so I can keep going with the material haha

Thanks for all the work you put on this, Mine!

Besides my comments above, the code LGTM!

[minestarks]

nit: should "agents", "prompts", "skills", all be in a folder centralizing all ai-related files? Just to make sure that in the future it's easy to find all of the md files and modify their behavior if necessary.

Staged at #3242

[minestarks]

Some of copilot thinking mentions the tool calls and internal things that the user doesn't really need to see. Would be nice if there was a way to disable showing that, the experience would be more fluid

I played around with this, and found that the user can control how the Thinking parts are displayed, not us, and the setting applies to all agents, not just us.

So I'm gonna call this a limitation of us being in VS Code - we'll have to write our own application if we want more control over chat behavior :/