Review comment redesign: audit + overhaul signals, dropdowns, content density, and visual design

[JSONbored]

Why

The public PR review comment (buildPublicPrIntelligenceComment / the unified-comment bridge in src/signals/engine.ts) has grown incrementally, feature by feature, over many PRs -- each addition made sense in isolation, but the whole has never been looked at as ONE thing a maintainer actually reads. Concrete signals it's due for a real pass:

• The main signals table mixes signals maintainers actually act on (linked issue, code review, related work, Gittensor context) with ones that get skimmed past (change scope, validation posture, contributor workload) -- confirmed by direct maintainer feedback, not a guess.
• "Contributor next steps" (publicSafeNextSteps in src/signals/engine.ts) is, on inspection, just a deduped re-listing of the SAME action text already shown in the signals table's own Action column plus preflight-finding actions -- it adds no new information, just repeats what's above it.
• The "Contributor context" row (contributorContextPanelResult) renders a red ❌ for any non-Gittensor contributor today, even though its own evidence text says "not a blocker" -- the visual and the semantics disagree.
• The "Improvement" row (improvementEvidenceText, PR panel: new advisory improvement-signal section #4744/Maintainer risk x value triage view #4745) concatenates up to 2 raw finding sentences plus the FULL LLM rationale sentence into one table cell -- it reads as a wall of text instead of a quick signal.
• The new "[BETA] Chat with Gittensory" collapsible (Add a beta usage dropdown for chat/ask commands to the PR review comment #5078) looks visually identical to every stable, long-shipped collapsible above it -- nothing distinguishes an experimental feature from a mature one except three characters of text in the title.
• No mechanism exists yet for turning a held/blocked PR's findings into something a contributor can paste straight into their own AI coding tool to fix -- CodeRabbit ships exactly this ("🤖 Prompt for AI Agents" per-finding, plus a newer "Fix All Issues" aggregate prompt), and it's worth deliberately evaluating for this product.

Scope

This epic is audit + design + prototype work across 6 sub-issues. None of them are "just ship it" -- each sub-issue's job is to research the current behavior, propose a concrete redesign with real before/after content, and implement the parts that are unambiguous, while flagging genuinely open design questions back to the maintainer rather than guessing on subjective calls (visual style, information density, exact copy).

• Visually separate BETA/experimental features from stable ones (Visually distinguish BETA/experimental review-comment features from stable ones #5096)
• Audit and redesign (or cut) "Contributor next steps" (Audit and redesign (or cut) the 'Contributor next steps' collapsible #5097)
• Slim and redesign the main signals table, including the Gittensor-context row + a new "Points Forecast" exploration (Slim and redesign the main PR signals table (cut low-value rows, redesign Gittensor context, explore Points Forecast) #5100)
• Simplify the "Improvement" signal row to a terse score/rating (Simplify the 'Improvement' signal row to a terse score instead of a text wall #5101)
• Research + prototype AI-agent-ready copy-paste fix suggestions (CodeRabbit-inspired) (Research + prototype AI-agent-ready copy-paste fix suggestions (CodeRabbit-inspired) #5102)
• Overall visual/information-design pass: less text-heavy, clearer hierarchy, more distinctive styling (Overall visual and information-density redesign of the PR review comment #5103)

Constraints (apply to every sub-issue)

• Public-safe: never introduce reward/payout/score/ranking/trust-score wording into the public comment (the existing sanitizePublicComment / containsPrivatePublicTerm forbidden-term guards must keep catching everything this epic adds).
• Config-as-code: any new behavior toggle is a .gittensory.yml field with a sensible default, not a hardcoded product decision -- matching every other feature this session.
• Byte-identical-by-default where reasonable: a repo that hasn't opted into a NEW capability shouldn't see its existing comment change shape.
• No dashboard/apps/gittensory-ui/ work is in scope here -- this epic is the GitHub-native PR comment only.

Non-goals

• Rebuilding the underlying signal-computation logic (readiness score, preflight findings, etc.) -- this epic is about what's SURFACED and HOW, not the scoring math itself, except where a sub-issue explicitly proposes replacing a low-value signal with a better one.
• Any change to gate/merge semantics -- this is a presentation-layer epic; the gate's own advisory/blocking behavior does not change.

[linear-code]

JSO-3666