Feature Request: Enable AI Agent Self-Introspection Commands

## Problem Statement

AI agents cannot execute Q Chat's built-in introspection commands (`/mcp list`, `/tools`, `/context status`), forcing inefficient workarounds for troubleshooting and capability discovery.

## Current Workarounds

**Manual workaround:**
1. User opens separate Q Chat session
2. Manually runs commands and reports back

**Programmatic workaround:**
```
User: "Check what MCP tools are available"
AI: [opens new terminal] → q chat
AI: [executes /mcp list in new session]
AI: "Here's what I found..."
```

**Why workarounds aren't ideal:**
- **Breaks conversation flow** - context switching disrupts problem-solving
- **Resource overhead** - spawning new Q Chat processes
- **Complexity** - multi-step process for simple introspection
- **Developer friction** - slows down debugging and development workflows

## Proposed Solution

Enable direct introspection within the same conversation:
```
User: "My MCP tool isn't working"
AI: [executes /mcp list directly] → "I can see your toolkit MCP server is disconnected..."
```

## Impact

- **Streamlined developer experience** - no context switching
- **Faster debugging** - immediate diagnosis vs multi-step workarounds
- **Better conversation flow** - seamless troubleshooting
- **Resource efficiency** - single process vs multiple instances

## Scope

**Proposed approach:**
- Enable AI execution of **all read-only slash commands** that provide system introspection
- If implementation complexity requires it, start with core commands (`/mcp list`, `/tools`, `/context`) and expand iteratively

**Security principle:**
- Only commands that read current state without modification
- No configuration changes or destructive operations

## Related Issues

Complements but differs from:
- #2840 (user tool management)
- #2058 (raw output viewing)

This focuses specifically on **AI agent autonomy** for self-diagnosis.

Das Issue nennt keine Dateien, Tests oder einen Einstiegspunkt. Beginne damit, die Verarbeitung von Slash-Befehlen und den Ausführungspfad des AI agent zu lokalisieren, und ermittle dann, wie `/mcp list`, `/tools` und `/context status` in der aktuellen Konversation ausgeführt werden könnten, ohne einen weiteren Q Chat-Prozess zu starten; als abgeschlossen gilt die Aufgabe, wenn schreibgeschütztes Introspektionsverhalten funktioniert, während konfigurationsändernde und destruktive Befehle weiterhin nicht verfügbar sind.

Vom Indexierungsmodell aus dem Issue-Text verfasst.