Forcefully shut down the runtime before existing the main task

[chkeita]

Summary of the Pull Request

The default behavior or the runtime is wait for all spawned work to be done before exiting. This was causing the task to hang when exiting.
This change adds a force shutdown at the end of the main function of the task.

closes #3377

[tevoinea]

Just to help me better understand:

• We'll never call rt.shutdown_background() until let result = rt.block_on(run(matches)); completes
• run(matches) will only complete when one of the futures complete from this select! macro

  onefuzz/src/agent/onefuzz-task/src/managed/cmd.rs

  Lines 101 to 114 in bcf3cf9

  	let result = tokio::select! {
  	result = config.run() => result,
  	// Ignore this task if it returns due to a querying error.
  	Ok(oom) = check_oom => {
  	// Convert the OOM notification to an error, so we can log it below.
  	let err = format_err!("out of memory: {} bytes available, {} required", oom.available_bytes, oom.min_bytes);
  	Err(err)
  	},
  	_shutdown = shutdown_listener => {
  	Ok(())
  	}
  	};

• select! claims to cancel the remaining branches

Waits on multiple concurrent branches, returning when the first branch completes, cancelling the remaining branches.

Does that mean the cancellation from select! happens in the background? And that prior to this PR, we wait for the cancellation to complete when we drop the runtime?

[chkeita]

Does that mean the cancellation from select! happens in the background? And that prior to this PR, we wait for the cancellation to complete when we drop the runtime?

This is not necessarily a problem with select!. The tokio docs says it waits on all the spawned work. Select will cancel everything related the task itself But something was spawned inside that task of it becomes an independent branch that is not waited on.