Monitoring should use Rollover API

[elasticmachine]

Original comment by @skearns64:

Today, we create daily indexes for ES and Kibana, but this is often wasteful, as these tend to end up as rather small indexes/shards that take up unnecessary cluster resources to manage. Using the rollover API, we could more easily specify a reasonable target number of docs that, when exceeded, would cause a new index to be created.

Using the rollover API does have the drawback in that it is less efficient to maintain a fixed history duration, so we'd need to decide if we were comfortable with delete-by-query before undertaking this.

[jakelandis]

closing in favor of #38470