Feature request: Ability to override global 'check_update_interval' per health check

The reason for a long check_update_interval is to avoid lots of writes to the cluster when a health check's output changes despite no change in state. But if a health check's output does not change often, it still suffers the long update lag time (5m by default).

When registering a health check, a check_update_interval parameter should be able to be supplied such that the health check's output is more quickly broadcast when it changes. It is not hard to conceive of a health check that only occasionally changes its output, and yet that change is important and should be broadcast out to the cluster right away. Prioviding per-health-check update intervals would allow for this behavior and many others.