Enable an endless quint run for long-running experiments

[bugarela]

I found myself having to estimate how long a command would take in order to have it run overnight and have results ready in the morning. I miss-estimated several times, probably due to using parallel and the scheduler being unpredictable. The problem was, I had no report of progress, since I was using parallel, and would have to start from scratch if I cancelled the command.

A workaround I found after struggling is to use the --ungroup flag from parallel and be able to see a buggy progress bar being updated by all processes at the same time. I also added some console.log statements to Quint every 1k samples so I had a concrete number from all processes instead of only the progress bar estimation.

This workaround enabled me to run much more experiments than what I'd have run with my initial estimated approach - also because I can't always estimate when I'll need to stop the experiment, at least in my life/setup context here.

Ideally, we should have a quint run --endless or something that keeps running simulation until it is told to stop, and either periodically or when stopped, produces a report (see #1577) that can be used to understand how many simulations were run.