Support Redis Cluster in the websocket server

Describe the feature you'd like

Let REDIS_URL name several seed nodes so the websocket server can use a Redis Cluster, the way redis.cluster already works on the Nextcloud side.

Why

The README recommends Redis for multi-node websocket deployments, and states that "Redis is required for multi-node clusters". On an installation that already runs Redis Cluster, that recommendation currently cannot be followed: RedisAdapter.createRedisClient() builds a single-node client with createClient(), which does not follow MOVED redirections.

The result is not a clean failure at startup. The server connects, the Socket.IO Redis Streams adapter sets up, and then roughly half of all operations fail at run time depending on which slot a key hashes to.

Measured against a 3-master / 3-replica cluster, writing 40 keys through createClient() pointed at one node:

successful: 17/40
failed    : 23/40
example   : MOVED 8308 127.0.0.1:7002

Nextcloud server itself supports Redis Cluster ('redis.cluster' => ['seeds' => [...]]), so an administrator who has already built one reasonably expects the whiteboard backend to use it. Today the only options are to stand up a second, non-clustered Redis just for whiteboard, or to give up multi-node.

We run a Nextcloud installation with a six node Redis Cluster and four websocket nodes behind a load balancer. We ended up deploying a separate single Redis instance for whiteboard, which is now the only non-redundant component in an otherwise redundant setup.

Proposed shape

Reuse REDIS_URL and treat a comma separated value as a list of cluster seed nodes:

STORAGE_STRATEGY=redis
REDIS_URL=redis://node1:6379,redis://node2:6379,redis://node3:6379

createCluster() is already available in the redis package the project depends on, so this is a small change and needs no new dependency. Credentials given on the first seed have to be applied to the discovered nodes as well, since cluster discovery reports them without auth.

One thing worth documenting alongside it: Redis Cluster only has database 0, so the /database_number suffix that works for a single node cannot be used to separate whiteboard's keys from other users of the same cluster.

I have this working and tested against a real cluster and will open a pull request.

Start at RedisAdapter.createRedisClient() and the redis package's createCluster() support, then trace how REDIS_URL is parsed. Done means comma-separated seed URLs work with discovered-node authentication, while the README documents Redis Cluster's database-0 limitation and the single-node suffix restriction; verify against a cluster.