Production deploy for musicshare.binarygnome.com

- docker-compose.prod.yml overlay: Caddy with automatic HTTPS on one
  origin (web at /, API at /api incl. websocket), internal-only Postgres/
  Redis/API/web, secure cookies, private-network servers blocked, nightly
  pg_dump backups kept 14 days
- deploy/export-db.sh and import-db.sh to carry a database to the server
  (same TOKEN_ENCRYPTION_KEY required for stored connections)
- Base compose no longer requires apps/api/.env
- README: deployment, Spotify via scrobblers, SMTP settings

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MzFywujuH9TiuXbr9vjoX8

deploy/import-db.sh

@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+# Restore a dump from export-db.sh on the server, replacing what's there.
+#   deploy/import-db.sh <dump-file>
+set -euo pipefail
+cd "$(dirname "$0")/.."
+file="${1:?usage: deploy/import-db.sh <dump-file>}"
+compose=(docker compose -f docker-compose.yml -f docker-compose.prod.yml --env-file deploy/production.env)
+
+echo "This REPLACES the server's database with $file."
+read -r -p "Type 'replace' to continue: " answer
+[ "$answer" = "replace" ] || { echo "Aborted."; exit 1; }
+
+"${compose[@]}" stop api worker web
+"${compose[@]}" up -d postgres
+until "${compose[@]}" exec -T postgres pg_isready -U musicshare -d musicshare >/dev/null 2>&1; do sleep 1; done
+"${compose[@]}" exec -T postgres pg_restore -U musicshare -d musicshare --clean --if-exists --no-owner < "$file"
+# Apply any migrations newer than the dump, then bring everything back.
+"${compose[@]}" run --rm migrate
+"${compose[@]}" up -d
+echo "Restored. Everyone needs to sign in again (sessions are per-domain), and Plex/Jellyfin webhook URLs now use the new domain."