[cmd] Fix RepeatCommand calling end() twice

[KangarooKoala]

If the command wrapped by a RepeatCommand finished naturally at the same time the repeat command was interrupted, .end() would be called twice (first .end(false) and then .end(true)).
This was especially problematic if the wrapped command is a ProxyCommand, since the first call to .end() would set its m_command to null and then the second call would attempt to call m_command.cancel(), resulting in a NPE.

[KangarooKoala]

Should we add tests for this behavior?

[Starlight220]

If the command wrapped by a RepeatCommand finished naturally at the same time the repeat command was interrupted

How would this happen?
You're right, due to #4706 the wrapped command is in an inactive state between finishing and RepeatCommand's next iteration, and thus calling end twice is UB.
This patch does indeed fix the issue; just add an explanatory comment and tests.

[KangarooKoala]

#4576 is about PS4 triangle methods, did you mean a different PR?
Regardless, I added the comment and the tests, which hopefully help explain how this can happen.

[Starlight220]

My mistake, I meant #4706.

[KangarooKoala]

There's a few other tests that use CommandScheduler.getInstance() instead of new CommandScheduler() in the try-with-resources, should I fix those as well in here or in a separate PR?

[Starlight220]

Don't fix them here.

[KangarooKoala]

Okay, any other changes to make?

[Starlight220]

(Assuming CI passes)
Looks good to me

[KangarooKoala]

This is what failed, and it doesn't seem to be related to this PR?

[KangarooKoala]

CI passes