fix(healthcheck) start a single timer for all workers

[locao]

Before creating a new timer, check if this Nginx instance is already running an active health check in any worker.

Fixes: Kong/kong#6788

[kikito]

This change seems ok to me, but some tests are red.

[Tieske]

I do not think this is correct. We need to bring back the get_periodic_lock function that dissapeared from the code (haven't checked when and where). But each HC needs one for healthy and one for unhealthy. Please check history.

The lock is an SHM key created with TTL == interval time, every worker tries to get /set that key, on success, they run the HC, on failure (already exists) another worker is already running the HC, and this worker should skip.

[locao]

Supersededby #67