feat: Chat moderation, rate limiting

[Harxhit]

Goal

Protect the messaging tier with Redis token-bucket rate limits and a moderation
pipeline that enforces bans/mutes/deletes instantly across pods.

Why this matters

§10: token-bucket per user/IP on chat, reactions, join, transport-create; moderation
filters chat ingest and fans actions out via Redis so all gateways enforce instantly.

Tasks

• Implement a Redis token-bucket limiter; apply to chat:message, reaction, joinRoom, transport-create
• Reject over-limit actions with backoff feedback to the client
• Add the moderation filter at chat ingest (profanity/spam/ban check) → fill the Day 1 hook
• Create moderation_events model per §7; record mute/ban/delete/timeout/flag
• Moderator actions PUBLISH to a control channel so all gateways enforce immediately
• Enforce ban list at join (ties to Week 2 Day 5 verifyViewerAccess)

Acceptance criteria

• Spamming chat/reactions trips the rate limit and is rejected with backoff
• A banned/muted user is blocked across all pods within moments of the action
• Moderation actions are recorded in moderation_events

Notes

Rate limits are also a DoS control (§10) — tune buckets per action class.

[Harxhit]

Removing transport create as creation of transport is an internal process joinRoom can handle it.