Add timestamp to rmw_publish tracepoint

[cwecht]

Depends on ros2/ros2_tracing#74
For some background: ros2/ros2_tracing#73

Question: Has write_w_timestamp + Time_t::now any drawbacks if compared with using plain write?

[MiguelCompany]

@cwecht Functionally, there are no drawbacks.
Semantically, there might be one issue if we want the tracepoint to be as close to the network/transport as possible (as stated in ros2/ros2_tracing#73 (comment))

Depending on the QoS settings, there might be a non-negligible blocking of the calling thread before the sample is added to the writer's history. Calling write will set source_timestamp to the time after that block

[cwecht]

I think the proximity of the tracepoint to the actual write call is not that important in this case, as we can not get any closer to the actual write operation right? For the tracing itself the timestamp passed here is not that important.
For a more precise tracing, adding tracepoints to the DDS implementation itself would be necessary, I suppose, right?

However, my original question was about using that approach having any significant impact on the behaviour of the RMW/DDS layer itself. But I take from you anwser @MiguelCompany that there is no differing behaviour to be expected, right?

[MiguelCompany]

there is no differing behaviour to be expected, right?

Correct. No differing behaviour other than a different source_timestamp being notified on the subscription.

[fujitatomoya]

lgtm