Function retryingFetch

Smart fetch with immediate failover and per-node health tracking.

For read operations:

• On failure, immediately try the next healthy node (no backoff within a round)
• After trying all nodes once (one round), apply backoff before the next round
• Stop after failoverThreshold rounds

For broadcast operations:

• Only retry on pre-connection errors (ECONNREFUSED, ENOTFOUND, etc.) where we know the request never reached the server
• NEVER retry after timeout or response errors to prevent double-broadcasting