fix: insufficient error handling for cluster down

[supercaracal]

The following fix was insufficient.

• fix: ensure recoverability from cluster state changes #379
• fix: ensure recoverability from cluster down for pubsub and transaction #381
• fix: ensure recoverability for pubsub #383

Also the following issue is related.

• Client is raising Redis::Cluster::NodeMightBeDown when Redis Cluster is observed to be healthy #368

[supercaracal]

  4) Error:
RedisClient::TestCluster::ScaleReadRandomWithPrimary#test_pubsub_with_wrong_command:
RedisClient::Cluster::ErrorCollection: 172.18.0.4:6379: (RedisClient::ReadTimeoutError) Waited 6.1 seconds
    lib/redis_client/cluster/node.rb:270:in `call_multiple_nodes!'
    lib/redis_client/cluster/node.rb:138:in `call_primaries'
    lib/redis_client/cluster/router.rb:255:in `send_wait_command'
    lib/redis_client/cluster/router.rb:41:in `send_command'
    lib/redis_client/cluster.rb:56:in `blocking_call'
    test/redis_client/test_cluster.rb:893:in `block in wait_for_replication'
    test/testing_helper.rb:35:in `swap_timeout'
    test/redis_client/test_cluster.rb:892:in `wait_for_replication'
    test/redis_client/test_cluster.rb:20:in `teardown'

known, flaky in jruby