Implement infinite loops, deadlocks, and memory leaks resilience for node workers. #222
Description
The current implementation doesn't support workers' isolation from each other. If some of the workers take too much CPU time and RAM for rendering a single request, it must get killed and respawned.