Add healthchecking

[hackaugusto]

There are multiple ways and purposes of doing healthchecks. I think we need two types:

• Health check to be used by a process manager, that can restart the process if it is unhealthy
• Health check for an external system, that can collect metrics and alert when the system is misbehaving

The second case should be based on metrics #144, either because the system is not replying, or some important metric is not changing (e.g. the block number for the store and block producer).

The first case, should have native implementations, namely:

• sdnotify for systemd
• a control plane command for docker [related: add control plane #83].
  • IMO it makes sense to use this strategy
• and maybe an http interface for third-party process managers (probably doesn't have to be done now)

[Mirko-von-Leipzig]

Implemented indirectly as a result of #675 - we use trace information in the monitoring backend to trigger health warnings and alerts.