Fix main loop not ending properly

[juliangruber]

There are some issues with the zinnia loop end logic. See for example these logs from core-fly:

Mar 26 15:03:07 core-fly [spark]  ⇣ checking for updates
Mar 26 15:03:08 core-fly [voyager]  ⇣ checking for updates
Mar 26 15:03:08 core-fly [spark]  ⇣ checking for updates
Mar 26 15:03:08 core-fly [voyager]  ⇣ checking for updates
Mar 26 15:03:08 core-fly [voyager]  ✓ no update available
Mar 26 15:03:08 core-fly [spark]  ✓ no update available
Mar 26 15:03:08 core-fly [spark]  ✓ no update available
Mar 26 15:03:08 core-fly [voyager]  ✓ no update available

We can see that 2 update loops were running at once, which means that in a previous run() invocation the loop should have been aborted, but wasn't.

This PR fixes this by only using one method: controller.abort() for aborting, and controller.signal.aborted for checking if it was aborted. Previously, shouldRestart.get() was used in conjunction with the AbortController, and in some places only one was checked or set.

Should fix issues like https://filecoinproject.slack.com/archives/C03S6LXSRB8/p1711639505087089. This has also been observed on core-fly:

Mar 26 15:13:09 core-fly [3/26/2024, 2:13:09 PM] INFO  Updated Zinnia module source code, restarting...

(no logs following)

[juliangruber]

I could remove this condition because here we know we always want to restart

[bajtos]

nice cleanup!