Add comment value / relevance to Comment Moderation experiment #514

What problem does this address?

The new Comment Moderation experiment introduced sentiment analysis and toxicity for comments, along with extensibility hooks around the analysis pipeline. During review discussion in #155, additional interest emerged around evaluating the overall value or relevance of a comment, not just whether it is toxic or negative. This could help identify low-value comments such as:

• spammy or engagement-bait replies
• off-topic responses
• generic “thanks” / "+1" / low-context comments
• comments that do not meaningfully contribute to discussion

In being able to detect those types of comments, it could support functions like:

• moderation workflows
• prioritization for community managers
• surfacing high-quality discussion
• future automation around comment handling

What is your proposed solution?

Update the Comment Moderation experiment to add a "Comment Value / Relevance" analysis feature that builds on the existing comment analysis infrastructure. Related discussion in #155 noted that some extensibility hooks already exist around system instructions and response schema, though additional JS extensibility may be needed for the full admin experience.

Potential capabilities:

• Analyze comments for relevance and contribution quality
• Store a normalized “value” or “relevance” score
• Add a dedicated column in the Comments screen and pill on the Activity dashboard widget
• Allow bulk processing similar to sentiment/toxicity analysis
• Optionally flag low-value comments for moderation or review
• Provide filters/hooks for custom scoring logic and thresholds