Fix action cancellation by passing status from JSON (backport #953)

[sea-bass]

Backports #953 to the humble branch.

[sea-bass]

I think the failing tests here may require some of the other backports to be in first.

[MatthijsBurgh]

Approval based on the diff being the same as the original PR. I only approve merging this PR, when the CI has been fixed (by merging other backports first).

[sea-bass]

Approval based on the diff being the same as the original PR. I only approve merging this PR, when the CI has been fixed (by merging other backports first).

The other backports did not fix the failing test, but also

• The test has nothing do with this diff
• I built this branch locally on Rolling and it passes

I'm not sure why, but I've gone ahead and disabled the test. Let me know if this is okay.

[MatthijsBurgh]

It looks like we have flaky tests

[sea-bass]

It looks like we have flaky tests

I'd call this something different than flaky if they consistently pass on Iron/Jazzy/Rolling, but seem to fail on Humble. But it's an issue nonetheless.

The only recent change that touched these files is #958 (which was part of these backports), but it's also strange that the failures would then happen on this (unrelated) PR instead of the backport itself.

Let me try revert this change and see what CI thinks.

EDIT: Nope, reverting that PR did not help.

[sea-bass]

Mysteriously, I pulled down a ROS 2 Humble Docker container and tried to run the failing tests I skipped in there... and they passed.