Timeout Threshold for SIGTERM

[nahmed3536]

Summary

Currently, the code use SIGTERM to kill all the children processes. This is the [ideal way to kill a process because it can be blocked or handled in various ways](https://komodor.com/learn/what-is-sigkill-signal-9-fast-termination-of-linux-containers/#:~:text=SIGKILL%20(also%20known%20as%20Unix,or%20handled%20in%20various%20ways.). However, an issue with SIGTERM I've noticed is that it can hang - the children processes are never killed and the code gets stuck waiting forever. In that situation, we'd want to use a SIGKILL which will force the process to terminate.

In this PR, I've added logic that will wait some amount time for SIGTERM to kill the process, then after the timeout threshold is exceed, it will kill any remaining / lingering jobs with SIGKILL. I've also made the timeout threshold a parameter in the CLI so users can customize the timeout.

Usage

1. Can run GPU Burn as is without specifying a timeout threshold time for SIGTERM

./gpu_burn

By default, the timeout threshold is 30 seconds. I selected this number because I ran GPU Burn several times with different durations and found 30 seconds was about the average time it took for the children processes to exit gracefully with SIGTERM when no issues were present (i.e., didn't hang).

2. Can exclusively use SIGKILL by specifying a very small timeout threshold

./gpu_burn -stts 0

In this case, it will jump to SIGKILL immediately since the timeout threshold is 0 seconds.

3. Can exclusively use SIGTERM by specifying a reasonable time window

./gpu_burn -stts 120

In this case, it will jump to SIGKILL after 120 seconds / 2 minutes, which is a reasonable upper bound for SIGTERM to do it's jump (based on my empirical evidence).

Test Plan

Run the above examples - works as expected.

[wilicc]

Yeah why not. It's tricky to kill ongoing GPU work but shouldn't break anything.