Finish GraphCache implementation for Pub/Sub

[Yadunund]

This PR:

• Refactors how we store and update the GraphCache within each session. Specifically we move away from a YAML::Node to an unordered_map (although implementing a conversion from our custom map to YAML may be helpful for logging or transmitting the graph as a serialized YAML over the network).
• Makes the switch to rely on liveliness tokens to broadcast changes to the ROS graph. The previous implementation of relying on put/del/get convey the changes is retained but commented out as it could be reused to support discovery with zenoh-pico.
• Implements some basic topic/node introspection functions in rmw (the rest will be done in a separate PR).

Note: Without the router, nodes will not be able to discover each other since multicast discovery is disabled by default in the node's session config. Instead, nodes will receive discovery information about other peers via the router's gossip function.

[francocipollone]

Great Job @Yadunund 🚀
I left a bunch of not-game-changing comments, ptal

[francocipollone]

nit: use find_if here and there

Suggested change

	auto erase_it = found_node->pubs.begin();
	for (; erase_it != found_node->pubs.end(); ++erase_it) {
	const auto & pub = *erase_it;
	if (pub.topic == node->pubs.at(0).topic &&
	pub.type == node->pubs.at(0).type &&
	pub.qos == node->pubs.at(0).qos)
	{
	break;
	}
	}
	auto erase_it = std::find_if(found_node->pubs.begin(), found_node->pubs.end(), [&node](const auto &pub) {
	return pub.topic == node->pubs.at(0).topic &&
	pub.type == node->pubs.at(0).type &&
	pub.qos == node->pubs.at(0).qos;
	});

Then proceed as you do:

        if (erase_it != found_node->pubs.end()) {
            // Found a matching element, erase it
            found_node->pubs.erase(erase_it);
            //...
            //...
            //...
        }

[clalancette]

I ended up rebasing this on the latest, as well as fixing some bugs that I found. I also addressed most of @francocipollone 's comments, with the exception of a switch to find_if. I'll take a closer look at that later.

However, with all of that done, ros2 node list and ros2 topic list -t both work entirely correctly for me. There is still an open TODO in there where a topic with multiple types on it won't be shown properly. To fix that, we'll need to revamp the way some of the data is put together. That's a project for next week.

[francocipollone]

I ended up rebasing this on the latest, as well as fixing some bugs that I found. I also addressed most of @francocipollone 's comments, with the exception of a switch to find_if. I'll take a closer look at that later.

However, with all of that done, ros2 node list and ros2 topic list -t both work entirely correctly for me. There is still an open TODO in there where a topic with multiple types on it won't be shown properly. To fix that, we'll need to revamp the way some of the data is put together. That's a project for next week.

Confirmed, it is working correctly. It seems it was my bad: I had some issues with the daemon. Basically, it was already running (I was working on another branch) and then when I checkouted this branch and built I hadn't stopped the daemon.

[Yadunund]

Thanks for cleaning this up Chris! I'll work on massaging the cache data structures to support multiple types on the same topic in a separate PR to keep the diffs minimal. If both of you are okay with the code here, should we merge this in first?

[clalancette]

Thanks for cleaning this up Chris! I'll work on massaging the cache data structures to support multiple types on the same topic in a separate PR to keep the diffs minimal. If both of you are okay with the code here, should we merge this in first?

Yeah, I think we should do that. The open issues here are 1) to switch to find_if as suggested by Franco, and 2) fix multiple types per topic. As long as we fix both of those in a follow-up, I'm good with merging this.