[dynamic infra] Implement the loader wachdogs to restart the services when stucked

The idea is to replicate the same kind of watchdog implemented today with systemd.

We should also check if the same issue exists in the dynamic infra

changed milestone to %Dynamic infrastructure [Roadmap - Tooling and infrastructure]

added kubernetes priority:Low labels

marked this issue as related to #4915

This [1] could possibly help in this situation too.

[1] swh/devel/swh-scheduler!369