Refresh stale health status on health check request

[zer0def]

This PR addresses the following failure scenario:

1. No Prometheus is periodically scraping this exporter, Patroni daemon is down
2. Exporter starts, during initialization attempts to connect to Patroni daemon, fails and caches the exception as 503 in self.collector.status.
3. Without the fix, any request to the /health endpoint will result in a 503 response, even if the Patroni daemon starts running properly, but only until the first metric scrape once Patroni is running.