feat: add base64 vision encoding for Local LLM agents

[Maurice-aXeTech]

fixes #2207

When using Ollama or other local LLMs with a vision-capable model, images uploaded to a bot DM were previously only sent as a file path (e.g. [User uploaded an image: /files/receipt.jpg]), which local models cannot fetch. This PR encodes images as base64 data URLs so vision models can actually read them.

Changes:

• ai.py: Added _build_vision_content() helper that base64-encodes images into OpenAI-compatible content blocks
• ai.py: Both image-handling paths in process_message_with_agent() now use vision content for Local LLM
• agents_integration.py: Fallback path now handles list content (vision) vs string content
• raven_bot.json + TypeScript types: Added enable_vision_base64 field (Check, default off)
• Bot settings → AI tab: Toggle to enable/disable vision encoding, only shown when Local LLM is selected

Before:

After:

[Maurice-aXeTech]

@nikkothari22 What do I need to do for a review of this PR?