Add ability to ask AI questions across multiple uploaded sources simultaneously

Currently, users can ask AI questions based on only one source at a time (PDF, Epub, link, etc.). This limits the application's potential for cross-document analysis and comparison.

**Feature request:**
Enable support for asking questions across multiple uploaded sources at once, similar to how it works in NotebookLM (where questions can reference an entire library of documents).

**Expected behavior:**
User selects 2 or more uploaded sources (files, links, etc.)

Enters a single question

AI analyzes all selected materials and returns a unified answer based on them

**Example user prompts:**
“Compare the approaches in these two PDFs”

“What are the key differences between these documents?”

“What common ideas are shared across the three uploaded files?”

Look at the existing single-source question-asking flow in the frontend codebase, likely in components handling file uploads and AI prompts. Identify how sources are selected and sent to the backend. The work involves modifying the UI to allow multi-select, changing the API call to send multiple source IDs, and understanding the backend's AI integration to ensure it can process multiple documents. Check for any existing state management for selected documents.

索引モデルが issue の本文から書いたものです。