feat: ephemeral channels with TTL-based auto-archiving

[baxen]

Summary

Adds ephemeral channels — channels that auto-archive after a configurable period of inactivity. Full vertical slice from schema through to the desktop UI.

Changes

Schema

• Add ttl_seconds (INT) and ttl_deadline (TIMESTAMPTZ) columns to channels table
• Add partial index idx_channels_ttl_expiry for efficient reaper queries

Database (sprout-db)

• Extend ChannelRecord with ttl_seconds and ttl_deadline fields
• Add bump_ttl_deadline() — resets the deadline on channel activity
• Add reap_expired_ephemeral_channels() — archives channels past their deadline
• Thread ttl_seconds through create_channel() and create_channel_with_id()

Relay (sprout-relay)

• Parse ttl tag from NIP-29 create-group events via shared resolve_ttl() helper
• Bump TTL deadline on every channel-scoped event (skip kind:9007 create)
• Add background reaper task (configurable interval via SPROUT_REAPER_INTERVAL_SECS, default 60s)
• Support SPROUT_EPHEMERAL_TTL_OVERRIDE env var for testing with short TTLs
• Emit channel_auto_archived system messages and update NIP-29 discovery events on auto-archive

Desktop

• Add "Ephemeral" checkbox to channel and forum creation forms (Zap icon, 1-day default)
• Thread ttl_seconds through Tauri commands and Nostr event builders
• Expose ttl_seconds and ttl_deadline in TypeScript types and API layer

Code Quality

• Extract shared resolve_ttl() in handlers/mod.rs (eliminates duplication between ingest.rs and side_effects.rs)
• Extract EPHEMERAL_TTL_SECONDS constant in AppSidebar.tsx

How It Works

1. Client creates a channel with a ttl tag (e.g., ["ttl", "86400"])
2. Relay stores ttl_seconds=86400 and sets ttl_deadline = NOW() + 86400s
3. Every new message in the channel resets ttl_deadline forward
4. Background reaper runs every 60s, archiving channels where ttl_deadline < NOW()
5. Archived channels get a system message and updated discovery events

Testing

• SPROUT_EPHEMERAL_TTL_OVERRIDE=60 — override all ephemeral TTLs to 60s
• SPROUT_REAPER_INTERVAL_SECS=5 — run reaper every 5s
• Both documented in .env.example

Multi-pod Safety

The reaper uses archived_at IS NULL as an idempotent guard — concurrent runs from multiple relay pods are harmless. A future pass will add pg_advisory_lock coordination alongside the workflow engine.