Watch memory and restart workers

[rybakit]

Consider to implement the "watch and restart workers" feature described in https://github.com/amphp/aerys/issues/135.

[kelunik]

We should only restart workers if there's no running task IMO.

[trowski]

New original issue link: amphp/http-server#135.

I wanted to gauge interest in this issue. It probably would be fairly straightforward to implement, but at the moment I don't see it as much of a priority. If a worker crashes, the worker is removed from the pool automatically. However this might cause a task to fail that had nothing to do with the reason for the worker crashing, so there is definitely reason to look at a soft memory limit and restart a worker gracefully before it hits that limit.

[rybakit]

@trowski I'm still interested in this feature for the reason you stated, though I can't say it's urgent or critical to me atm.

[Ekstazi]

+1