Fallback discovery and keep-alive commands do not consider cluster slots

[mgravell]

• when INFO is disabled, AutoConfigureAsync uses SET $uniqeid$ replica_read_only PX 1 as the fallback mechanism for detecting read-only replicas, without consideration of slots; whether this works correctly on OSS cluster is questionable, but: on OSS cluster we don't need to do this; we should respect the cluster topology data instead
• when ECHO, PING and TIME are disabled: GetTracerMessage uses EXISTS $uniqeid$ as the final fallback on interactive connections, without consideration of cluster slots; suggestion: on OSS cluster, use a hash-tag in the key

Generating a hash-tag for an arbitrary slot is relatively fast, and could be hard-coded or memoized, allowing for the key {$prefix$}$uniqueid$ (where {/} here denote hash-tag literals, not injection tokens) to be used - where prefix is for any slot (presumably the first, for simplicity) for that endpoint

Additionally: we probably shouldn't do anything with the tie-breaker key if we're on cluster.