Add RAG chatbot (Pixel) and admin config panel

[yiilinzhang]

Description

Adds a new RAG-powered course chatbot (Pixel) for Academy pages, an admin panel for configuring Pixelbot prompts.

Changes

Before	After
No course-level chatbot	Pixel chatbot on Academy pages with RAG-powered responses
No admin config for chatbot prompts	Pixelbot Settings tab in admin panel
Chatbot fixed position	Draggable chatbot with minimize/maximize
Plain text responses	Markdown-rendered responses (code, tables, lists)

Key changes

• src/commons/application/types/SessionTypes.ts — Added pixelbotRoutingPrompt, pixelbotAnswerPrompt, feedbackUrl to session and course config types
• src/commons/sagas/RequestsSaga.ts — Added getPixelbotDocumentMap API call
• src/features/ragChat/api.ts — New API layer for RAG chat (initRagChat, sendRagMessage)
• src/pages/sicp/subcomponents/chatbot/ChatBox.tsx — Added expand/collapse support
• src/pages/sicp/subcomponents/chatbot/Chatbot.tsx — Made draggable
• src/pages/academy/adminPanel/AdminPanel.tsx — Added Pixelbot Settings tab with new config fields
• src/pages/academy/adminPanel/subcomponents/PixelbotConfigPanel.tsx — New admin panel for routing/answer prompts and feedback URL
• src/pages/academy/ragChatbot/RagChatBox.tsx — New chat UI with markdown rendering for RAG responses

Type of change

• New feature (non-breaking change which adds functionality)

How to test

2. Log in and open Source Academy
3. Click on Pixel chatbot in bottom-right corner
4. Send messages to Pixel like "What was covered in week 1" and "Give me some past year practice questions" to test it out
5. Click on code blocks in a response to check out new playgrund integration
6. Go to Admin Panel → Pixelbot Settings — Edit routing prompt, answer prompt, document map, and feedback URL fields here

Checklist

• I have tested this code

[gemini-code-assist]

Code Review

This pull request introduces 'Pixelbot,' a RAG-based chatbot for the Academy platform, which includes a draggable chat interface, new API endpoints, and an admin configuration panel for managing prompts and feedback URLs. The existing SICP chatbot was also refactored to support expansion and draggability. Feedback focuses on improving Markdown rendering by utilizing ReactMarkdown components instead of manual regex parsing and preventing a potential infinite loop in the configuration panel by selecting individual token fields for the useEffect dependency array.