This PR is a smorgasbord of small fixes around the tree. It is probably easiest to look at the individual patches, but in short:

• Add a small optimization where we keep a running total of the number of nodes in the graph. This speeds up get_nodes_names for large graphs, since we don't have to iterate over the namespaces anymore.
• While doing the above, I noticed that we never removed namespaces once they were added. I added a patch so that if there are no nodes left in the namespace after this one is removed, we also remove the namespace.
• I removed a few old TODO comments.
• I removed the PublishToken implementation. While it is somewhat nice to have, it is a bit messy and I think we are going to have to do something more substantial when we try to support zenoh-pico. And this code is still in the git history.
• Remove some hand-coded rcutils_strdup implementations. The original reason for that hand-coding is gone, so we can just use rcutils_strdup now.
• Speed up ros_topic_name_to_zenoh_key. In particular, std::to_string is much faster than std::stringstream, so use that.
• Use RAII for saved_msg_data. That way we don't have to remember to z_drop the sample.
• Implement rmw_compare_gids_equal, which just gets us one step closer to having a complete implementation.