[Monitoring] More resilient against errors with alerting

Currently, if there is any kind of error with fetching alerts, the Stack Monitoring UI fails to load (as we don't explicitly catch any errors [when fetching alerts](https://github.com/elastic/kibana/blob/master/x-pack/plugins/monitoring/server/lib/cluster/get_clusters_from_request.js#L154) when loading all clusters) which shouldn't happen.