feat(analytics): add outcome learning loop for recommendation quality

[JSONbored]

Parent roadmap: #127
Parent phase: #237
Related advanced phase: #269
Related base feedback issue: #96

Goal

Add an outcome learning loop that measures whether recommendations are useful and feeds that signal back into future ranking quality.

This issue is now a mini-epic. Do not implement the full analytics loop in one PR; use the focused child issues below.

Child issues

• feat(analytics): persist recommendation outcome events #301 persist recommendation outcome events
• feat(analytics): classify recommendation outcomes #302 classify recommendation outcomes
• feat(analytics): add role-aware recommendation quality rollups #303 add role-aware recommendation quality rollups
• feat(agent): feed aggregate outcome quality into confidence #304 feed aggregate outcome quality into confidence
• feat(app): expose private recommendation quality report #305 expose private recommendation quality report

Done when

• Outcome events, classification, role-aware rollups, confidence feedback, and the private quality report are complete or intentionally superseded.
• Recommendation quality influences ranking through deterministic private signals.
• Public surfaces never expose private scoreability, reward/risk, wallet, hotkey, raw trust, or private reviewability.

Non-goals

• No public reward prediction or public scoring.
• No autonomous PR bot.
• No public leaderboard.

Validation expected

• Child PRs should include outcome fixtures, analytics rollup tests, sanitizer coverage, and npm run test:ci.