Add "watchdog" script on Frontier runs

[titoiride]

In this PR we add a script that checks periodically if the code execution on Frontier is proceeding as expected.
In case the code is hanging, which is verified by checking the output file timestamp, the execution is killed to save computational time.

Compared to previous attempts, this script has a double timer: a frequent check and an output modification check. The check is ran frequently (every check_interval seconds) and checks that the output file has been modified not more than timeout_sec seconds before. Differentiating the timers, compared to just running a single check, prevents the script to stall a simulation that, for instance, finishes correctly between two checks. This way, if the simulation ends between two timeout_sec checks, the script will wait at most check_interval before closing.

Close #5584