Improve cache refresh error handling for offline mirror nodes

[jbair06]

Description:
Cache refreshing doesn't handle offline/unavailable mirror nodes gracefully. When a mirror node goes down, the current error handling blocks processing and produces verbose/less-useful logs. The system should treat mirror node failures as non-blocking, log them succinctly, and potentially implement backoff per offending network.

Current behavior:

Mirror node X offline → Cache refresh blocks → Processing stalls → Verbose error spam in logs

Expected behavior:

Mirror node X offline → Non-blocking error → Succinct log entry → Continue processing other nodes
→ Optional: Backoff retry for network X specifically

Scope:

• Make mirror node failures non-blocking (continue processing other nodes)
• Improve error logging to be more succinct/actionable
• Track failures per network/mirror node for backoff logic
• Ensure cache refresh failures don't halt transaction processing

Acceptance criteria:

• Mirror node timeout/offline doesn't block cache refresh for other nodes
• Cache refresh failures are logged succinctly (network + node + error type)
• Processing continues immediately despite individual node failures
• Per-network backoff tracking prevents log spam from repeated failures
• Notification/messaging processing is never blocked by cache refresh errors

Nice to have:

• Circuit breaker per mirror node/network
• Metrics for mirror node health/failure rates