CachingResolver: stop retrying permanently failing hosts

## Problem When `expireAction == Refresh`, a host that fails DNS resolution is retried every `unknownHostTtl` (default 1 minute) indefinitely. Each retry consumes a semaphore permit and makes a blocking system DNS call. For hosts that will never resolve, this wastes resources forever. ## Proposal Add a max retry count to `CachingResolver`. After a configurable number of consecutive failed refreshes, the entry would be dropped from the cache instead of being retried. This would require: - A new `failureCount` field on `CacheEntry`, incremented in `refreshOrDropEntries` each time a failed entry is refreshed - A new `maxRetries` config field in `DnsResolver.Config` (with a sensible default) - When `failureCount >= maxRetries`, drop the entry regardless of `expireAction` ## Alternatives - **Max failure TTL**: Track when the host first started failing and drop after a duration ceiling. Similar complexity (new field on `CacheEntry`) but uses time rather than count. - **Exponential backoff**: Progressively increase the retry interval. Reduces resource usage without a hard cutoff, but adds complexity. @987Nabil Which solution do you prefer?