keeper-client commands are resource inefficient

### Issue

I tried using two subcommands of keeper-client on 90M znode cluster
- find_super_nodes
- find_big_family

Found that keeper memory spiked by more than 6GB on the server where the above commands were executed. This may cause bad impact on the production clusters since we have soft limit at 90% of memory.

Both the commands didn't complete i had to cancel the commands.

Following two spikes in the keeper memory are due to the above two command where none of them completed.

### Describe the improvement

These commands are used during critical times when something is already not right on the cluster. Any such commands run by users should take minimal resources or atleast warn users if they are going to use unexpected resources and shouldn't bring down the cluster.

### Which ClickHouse versions are affected?

I ran it on 25.10 but pretty sure latest versions are also impacted.

### How to reproduce

- Create a keeper cluster
- Generate lots of znodes
- Use the above two commands

### Error message and/or stacktrace

_No response_

### Additional context

_No response_