Fix interference between test_tracetools and ros2lifecycle tests

[christophebedard]

Part of #94

Relates to #95, but is still separate

test_tracetools's test_lifecycle_node test and ros2lifecycle's test_cli test can interfere with each other if run in parallel, making one of them -- or both of them -- fail. For example, test_tracetools fails in this nightly job here: https://ci.ros2.org/view/nightly/job/nightly_linux_debug/2988/.

This is mainly because both tests use a lifecycle node with the same name and without any specific namespace:

1. test_tracetools's test_lifecycle_node test uses test_lifecycle_node
2. ros2lifecycle's test_cli test uses ros2lifecycle_test_fixtures's simple_lifecycle_node, which is named test_lifecycle_node

This can lead to two separate failures:

1. Since the lifecycle nodes have the same name, without any specific namespace, the tests can end up triggering transitions for the wrong lifecycle node
   • I think this sometimes results in test_tracetools's test_lifecycle_node test timing out because the state machine ends up in a weird state (not sure why the timeouts in test_lifecycle_node.cpp don't make it exit before the 60-second timeout, though)
2. The trace that test_tracetools's test_lifecycle_node test collects can include transitions from ros2lifecycle

The two separate solutions are:

1. Use a namespace (/test_tracetools) for the node in test_tracetools's test_lifecycle_node test to avoid transition/topic-level interference
2. Make sure to only consider transition trace events for the right state machine instead of assuming that all transitions in the trace come from the expected state machine and then checking if they correspond to the expected state machine
   • This would also be fixed by Allow tracing tests to be run in parallel with other tests #95 even though they are different changes, but this change doesn't really hurt

[clalancette]

Looks good to me.

For CI, I think it might make sense to run CI tests both sequential and parallel, though I'll leave it up to you.

[christophebedard]

Yeah I can do that. Tracing tests might fail in parallel without #95, but this particular test should pass.

[christophebedard]

Sequential testing of tracetools_test/test_tracetools and above only:

• Linux
• Linux-aarch64
• Windows

Parallel testing of everything:

• Linux
• Linux-aarch64
• Windows

[christophebedard]

• Linux-aarch64

Those test failures are "expected" (see #95) and not related to this PR.

[christophebedard]

Alright, CI results are acceptable, I'll merge this and rebase #95.