FEATURE: Add AI-powered spam detection for new user posts

[SamSaffron]

This PR introduces an AI-powered spam detection system that helps protect communities from spam by automatically scanning posts from new users.

Key Features:

• Automatic scanning of the first 3 posts by new users (TL0-TL1)
• Smart edit detection to catch spam modifications
• Admin UI for configuration and monitoring
• Integration with existing spam handling mechanisms
• Statistics tracking for accuracy monitoring
• Custom instructions support for site-specific rules

Technical Implementation:

• New AiSpamLog and AiModerationSetting models for tracking and configuration
• Post-creation and post-edit hooks for automated scanning
• Intelligent rate limiting and edit detection
• Custom LLM prompt engineering for spam detection
• Integration with Discourse's existing spam handling system

Admin Features:

• Enable/disable spam detection
• Select LLM model for scanning
• Configure custom site-specific instructions
• Monitor detection statistics (scanned posts, detected spam, false positives/negatives)

Safety Features:

• Only scans public posts
• Limited to first 3 posts by new users
• Minimum edit difference threshold
• Rate limiting for rescanning edited posts
• Integration with existing trust system

[SamSaffron]

Left to do:

1. Some way to drill into the 4 lists (false positive / negative / scanned / spam)
2. Send images through

Except for that, this is pretty much ready

[keegangeorge]

Follow up changes look good, and overall everything looks good while testing locally. A few 🤏🏽 tiny things

handled all issues