Improved interrupt/cancelation build tests

What does this MR do?

• Introduces RunBuildWithCancel, a test that allows us to test whether interrupts and cancelations are working during a build.

Why was this MR needed?

• The logic was previously copied/pasted to all executors.
• Most tests were behaving incorrectly: A delay of 1 second, for most executors, was not enough to cancel a context during a build.
• We're going to introduce another form of cancelation, and we need to test the behaviours between them: on some, after_script should be executed, and on others the build should immediately terminate.

What's the best way to test this MR?

Each executor has a *BuildCancel test, eg:

go test -v -count 1 -run TestDockerCommandBuildCancel ./executors/docker

What are the relevant issue numbers?

#4843 (closed)