Add delay after node creation to ensure discovery for other rmw implementations

[BMarchi]

For connext, there were two issues:

• https://ci.ros2.org/job/ci_linux/8479/testReport/junit/sros2.test.sros2.commands.security.verbs/test_generate_policy/test_generate_policy_services/
• https://ci.ros2.org/job/ci_linux/8479/testReport/junit/sros2.test.sros2.commands.security.verbs/test_generate_policy/test_generate_policy_topics/

When the cli command is run,

sros2/sros2/test/sros2/commands/security/verbs/test_generate_policy.py

Lines 41 to 42 in c7ae64e

	assert cli.main(
	argv=['security', 'generate_policy', os.path.join(tmpdir, 'test-policy.xml')]) == 0

it runs in another thread/context, ending up here:

sros2/sros2/sros2/verb/generate_policy.py

Line 130 in c7ae64e

subscribe_topics = get_subscriber_info(node=node, node_name=node_name)

The culprit here, is the next portion of code:
https://github.com/ros2/rmw_connext/blob/b1062f92bfbec2e85be3083144a0f6c5a599287d/rmw_connext_shared_cpp/src/node_info_and_types.cpp#L104-L139

get_discovered_participants from connext is returning 0 handlers, concluding that connext is not finding the previously created node in python.
A timer is not ideal, but for testing purposes I see no other way to ensure the node's advertising.

[hidmic]

@BMarchi it's true that discovery latency makes it hard for us to write a robust test, but consider looping up to N times with a shorter timeout instead. It'll hopefully result in shorter run times.

[BMarchi]

I was checking your suggestion @hidmic, but sometimes the test fails, even with my changes. Since we have to make retries and specify longer timeouts for ros2cli tests for every rmw_implementation for example, shouldn't we do the same thing here since it's using a cli feature?

[BMarchi]

I changed the test to start using launch_testing instead, take a look whenever you can @jacobperron , @hidmic

[BMarchi]

• Linux
• Arch
• macOS
• Windows

[jacobperron]

One comment below, otherwise LGTM

[jacobperron]

I'm not sure that this PR resolves the previous test failure on Windows (#143).
We probably still want to keep the TODO and skip decorator. Note, to reproduce the failure, you should also run tests for packages other than sros2 (e.g. colcon test --packages-select demo_nodes_cpp && colcon test --packages-select sros2).

[BMarchi]

Re run CI for linux:

• Linux

• Windows

[hidmic]

LGTM but for a few nits.

[hidmic]

LGTM!

[hidmic]

Blocked on ros2/ros2cli#419.

Re-runing CI up to sros2:

• Linux
• Windows

[hidmic]

Rebased, refactored and used as a basis for #214. Closing.