Deprecate 'ctl node_health_check'

[michaelklishin]

• It requires a fully booted node, so not generally suitable for a Kubernetes readiness probe.
• It can produce false positives
• It is too intrusive and CPU-intensive to use at scale
• Most operators do not understand what it really does and when they learn about it, consider it to be too opinionated and intrusive

Time for the One True Health Check™ to retire from duty.

Part of rabbitmq/rabbitmq-cli#426

Per discussion with @lukebakken

[michaelklishin]

Backported to v3.8.x.