Automatically test a random DB for clustering tests

[madolson]

          A few lines below this we set `::singledb` to 1. Normally we test on a random DB, which has helped us catch some random regressions in the past, I think it would be helpful if we can remove that to improve coverage. I'm okay skipping that for now, but would like to follow up and add it.

Originally posted by @madolson in https://github.com/valkey-io/valkey/pull/1671#discussion_r2057181290