Support sampling/limiting topology refresh source nodes for large clusters

[dlgocks1]

Feature Request

Is your feature request related to a problem? Please describe

With dynamicRefreshSources(true), Lettuce queries all discovered nodes during topology refresh. In our 256-node cluster, this causes unnecessary load and connection churn.

Current options are all-or-nothing:

• true: Query all 256 nodes
• false: Query only initial seed nodes (loses dynamic discovery)

Related: #240, #1394

Describe the solution you'd like

Add option to limit/sample the number of nodes queried:

ClusterTopologyRefreshOptions.builder()
    .dynamicRefreshSources(true)
    .maxTopologyRefreshSources(5)  // Only query up to 5 nodes
    .build()

Describe alternatives you've considered

• dynamicRefreshSources(false): Works but loses dynamic node discovery
• Longer refresh intervals: Delays failover detection

Teachability, Documentation, Adoption, Migration Strategy

// For large clusters (100+ nodes), limit topology refresh sources
ClusterTopologyRefreshOptions.builder()
    .dynamicRefreshSources(true)
    .maxTopologyRefreshSources(10)
    .build()

[bandalgomsu]

Hi @tishun !!

I’m interested in this issue.

May I work on it?

[a-TODO-rov]

Hello @bandalgomsu
Of course, feel free to contribute to this one.

[bandalgomsu]

Open PR ! #3586